tag:blogger.com,1999:blog-305734582024-03-13T13:58:50.970-06:00Saiph's BlogCivic Tech, Machine Learning, Deutsch Musik, and Research...Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.comBlogger117125tag:blogger.com,1999:blog-30573458.post-74576601154142369502023-06-21T04:25:00.002-06:002023-06-21T04:25:41.732-06:00Into the World of AI for Good: Reflections on My First Week in the Civic AI Lab at Northeastern
By Undergraduate Researcher Liz Maylin <br><br>
<div class="separator" style="clear: both;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhyXt1evaWZrTo_MTFcLbH-XSjYGUtgtHYU6NmzzhkKzUD0_NLRuDRcFss5fIw6d5bw1BwYBKTMUFfpi8TWFiytFjZq4KfG7LZvrvZPT4J71mVJXMkh-7vIv0G4plMmUTlDHV7WL30pS5dr-0HQcDDFszOY47FHpeyca4Hyy4iuZZv7jNuXHFDJ/s512/liz.jpeg" style="display: block; padding: 1em 0; text-align: center; clear: left; float: left;"><img alt="" border="0" width="200" data-original-height="512" data-original-width="512" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhyXt1evaWZrTo_MTFcLbH-XSjYGUtgtHYU6NmzzhkKzUD0_NLRuDRcFss5fIw6d5bw1BwYBKTMUFfpi8TWFiytFjZq4KfG7LZvrvZPT4J71mVJXMkh-7vIv0G4plMmUTlDHV7WL30pS5dr-0HQcDDFszOY47FHpeyca4Hyy4iuZZv7jNuXHFDJ/s200/liz.jpeg"/></a></div>
Last Tuesday, I began my journey in the Civic A.I. Lab at Northeastern University with a mix of excitement, curiosity, and gratitude for the experience. A special thanks to Professor Eni Mustafaraj (Wellesley College) and Dr. Saiph Savage (Northeastern) for this opportunity. I embarked my first week in this transformative space, eager to learn and contribute to the lab, which studies problems involving people, worker collectives, and non-profit organizations to create systems with human centered designs to address these problems. Some of the objectives of the lab include fighting against disinformation and creating tools in collaboration with gig workers. Previous projects include designing tools for latina gig workers, systems for addressing data voids on social media, and a system for quantifying the invisible labor of crowd workers. Nestled at the crossroads of Human-Computer Interaction, Artificial Intelligence, and civic engagement, the research of this lab is thoughtful and resoundingly impactful. I am honored to join a project that will support workers in their collective bargaining efforts.
<div class="separator" style="clear: both;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjFziRBzOFre_pvUk7q3aW78ZG3knXzHyEP8C_KanY7C4bkztQEdbgQioHiVKWlNZNgsl916YN-pL5V0Tvb-gb6rSb74dChVoixdbJraD0vx9Q_vo4_vK10YLhf_EOXnWimJxxpDVTTbuTZCQVpePusFc59p2GZhZI5OsB6N0vAB1LOYAJdwO3W/s400/LowvphCD_400x400.jpg" style="display: block; padding: 1em 0; text-align: center; "><img alt="" border="0" width="320" data-original-height="400" data-original-width="400" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjFziRBzOFre_pvUk7q3aW78ZG3knXzHyEP8C_KanY7C4bkztQEdbgQioHiVKWlNZNgsl916YN-pL5V0Tvb-gb6rSb74dChVoixdbJraD0vx9Q_vo4_vK10YLhf_EOXnWimJxxpDVTTbuTZCQVpePusFc59p2GZhZI5OsB6N0vAB1LOYAJdwO3W/s320/LowvphCD_400x400.jpg"/></a></div>
<h3>First Impressions at Northeastern</h3>
During my first days, the differences between Wellesley College and Northeastern stood out to me the most. Wellesley College is located 12 miles outside of Boston in the extremely quiet, wealthy town of Wellesley whereas Northeastern is located directly in the city, allowing for greater access and a larger community. I get to walk past Fenway Park, various restaurants, boba shops, a beautiful park, and the Museum of Fine Arts on my commute! It is definitely a change of setting but I am happy for the experience. So much to explore! <br>
First weeks are exciting because there is so much to learn and new people to meet. The research team is full of amazing, talented students that I am excited to collaborate with and learn from. As I prepare for the application process for graduate school, I am fortunate to gain insight into the lives of PhD and Master’s students that will help me make informed decisions for my own academic journey. Everyone has been very welcoming and helpful, I am thrilled to spend this summer with them.
<h3>Exploring Gig Work and Participatory Design</h3>
I have spent most of my first week getting familiar with gig work and participatory design through literature review. Gig work is a type of employment arrangement where individuals perform short-term jobs or tasks. This work includes independent contractors, freelancers, and project based work. Often, gig work is presented as the opportunity to “be your own boss” and “ to work on your own time”, however this line of work comes with challenges such as irregular income, limited job security, and typically no benefits. The use of digital platforms has facilitated connection between workers and employers; however, there is room for improvement that will benefit both users and platforms. Participatory design is a method that includes stakeholders and end-users in the process of designing technologies with the goal of creating useful tools or improving existing ones. For example, researchers at the University of Texas at Austin held sessions with drivers from Uber and Lyft to reimagine a design of the platform that would center their well-being. It’s fascinating work that unveils different solutions and possibilities capable of reconciling stakeholder and worker issues.
<h3>Learning about Data Visualization</h3>
Additionally, I have been getting acquainted with different forms of data visualization. I have some experience programming with python but usually for problem sets or web scraping so I was filled with anticipation to acquire a new skill. Specifically, I have focused on working on text analysis. With the help of tutorials, google, and Viraj from our lab, I was able to make a wordcloud that showed the most frequent words in a dataset that included reviews of women’s clothes from 2019 (shown below).
<div class="separator" style="clear: both;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgJ7HP4EtAUvsN_6XpQSifXbcda8po3NpHj3r_c9PiPxGyzWJX-Nuz7iksDfo2rOd9unrEy6Z4qZ4A5kvw9xA_QiILL6LJs9AujNBRVwYexqYqdnOy_qVWv1csM-qDIYQ9T9hzApdQKrPuz-yvBOufkX2wxXEIXFP-OMAYRyr4E2w-ePnejCa1j/s586/kk.png" style="display: block; padding: 1em 0; text-align: center; "><img alt="" border="0" width="320" data-original-height="294" data-original-width="586" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgJ7HP4EtAUvsN_6XpQSifXbcda8po3NpHj3r_c9PiPxGyzWJX-Nuz7iksDfo2rOd9unrEy6Z4qZ4A5kvw9xA_QiILL6LJs9AujNBRVwYexqYqdnOy_qVWv1csM-qDIYQ9T9hzApdQKrPuz-yvBOufkX2wxXEIXFP-OMAYRyr4E2w-ePnejCa1j/s320/kk.png"/></a></div>
Through this process, I was able to learn about various resources such as Kaggle and datacamp that provide datasets and tutorials to practice working with data. I originally tried using the NLTK library but I had several problems with my IDE (VSCode). With some troubleshooting help from lab members, I switched my approach to just using pandas, matplotlib, and wordcloud. I am happy I got it working and I’m looking forward to refining this skill.
As I wrap my first week, I am beyond excited for the opportunities that lie ahead. This experience has ignited a passion for leveraging technology for civic engagement. I am grateful for the warm welcome, the technical help, and the inspiring conversations from this week. I am eager to collaborate and contribute to the work of the lab. :)
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-32424312544635134572023-06-20T18:20:00.000-06:002023-06-20T18:20:05.011-06:00Starting My Summer Internship at the Northeastern Civic AI Lab<h2 id="summer1">Starting My Summer Internship at the Northeastern Civic AI Lab</h2>
<div class="separator" style="clear: both;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhfiJwvgUbMl4WYnfB6TvjOCiVmRXlQTosjA_nNMnJq0Wc2JNbJNF9iMj-a1-5V9nr49SzUClAQ2K1tm2uhzwHNc-FyAajvDLbxHmJZk3iTL-infzrwHQjLxQ5S0Ia84H6zih3Cm0yz5Ce9ILC7oMknZ8BfmxY9o2yA-RK_DonaoWDJhMLZE9FJ/s176/Screenshot%20from%202023-06-20%2020-07-03.png" style="display: block; padding: 1em 0; text-align: center; clear: left; float: left;"><img alt="" border="0" height="200" data-original-height="176" data-original-width="141" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhfiJwvgUbMl4WYnfB6TvjOCiVmRXlQTosjA_nNMnJq0Wc2JNbJNF9iMj-a1-5V9nr49SzUClAQ2K1tm2uhzwHNc-FyAajvDLbxHmJZk3iTL-infzrwHQjLxQ5S0Ia84H6zih3Cm0yz5Ce9ILC7oMknZ8BfmxY9o2yA-RK_DonaoWDJhMLZE9FJ/s200/Screenshot%20from%202023-06-20%2020-07-03.png"/></a></div>
<h3>Intro</h3>
Hi! I’m Simon Juknelis, a rising high school senior at Noble and Greenough in Dedham, MA, and this week, I am beginning my work as an intern at the Civic AI Lab at Northeastern University. I’ve always been interested in building projects that help other people and make an impact, and that’s what I hope to achieve over the course of this internship.<br>
The lab’s overall mission is to build new technology solutions that create equitable positive impacts and that empower all members of society. To achieve this goal, the lab works with non-profit organizations such as the National Science Foundation and UNESCO as well as tech industry leaders such as Twitch and Meta. The lab has done research into a large number of areas such as preventing disinformation using AI and data labeling work, and I’m very excited to get to work with this team!
Participatory Design<br>
Our lab’s work over the coming months will involve building software solutions for use by and to benefit gig workers. As such, I read up on the methodological framework of participatory design. Participatory design encapsulates the idea of giving the users of a product the power to shape it to fit their needs. Participatory design can be carried out with interviews and workshop sessions with a sample of potential users of the product. The future users should be given the ability to give suggestions during the ideation phase of the product as well as at various stages throughout its development.
Our lab will be using participatory design over the next few months in order to conduct our research and build solutions. As we work on finding research participants and setting up interviews, I decided to test my technical skills by building a small web plugin called ProductWords, which allows users to look through Amazon products, add them to a list, and see statistics about them.
<h3>Participatory Design for Gig Workers</h3>
In fields like gig work, there can often be a large power imbalance between a single worker, on one hand, and the corporate clients and work platforms, on the other hand, that provide the gig worker’s income. As such, when tools are designed for gig workers, they are often designed without gig workers’ actual needs in mind, and instead are designed by the platforms based on what they think the workers need or even based on what would benefit them or the clients. As such, participatory design is an important tool to ensure that tools built for gig workers actually benefit those workers.
<h3>Designing Tools for Gig Workers with Figma</h3>
Part of our lab’s work will involve using a software platform called Figma for UI/UX design. One of the main benefits of using a service like Figma is the collaborative benefits it provides. It allows for ideas about interface layout, animations, and functionality to be more easily communicated between team members, and multiple team members can work together on the same files to create a unified design workspace.
I used Figma to design the interface for ProductWords. Doing it this way was especially helpful because I have not yet finished implementing all of these visual elements into ProductWords, but I still have a good sense of what I want the final product to look like and I’ll be able to look back on this Figma doc to see what I should implement.
<div class="separator" style="clear: both;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhbVxP9ODFkPXtjjfi8o48W8IyUNZJivfe0oU_y-GEiU4Ak8QRUDLxzdsKuhpf8UmzNF0aSvPBCW9APV3u-iUeYYlP9pXPi0FbaK7pGGFp4_Fe_MNYRMsdxMWQOc7uiT12cMWtvWAgAXNunOaKjJLB9cjWG3xcL0_8jZsQsIatCuPWskbFy1vG4/s582/Screenshot%20from%202023-06-20%2019-56-18.png" style="display: block; padding: 1em 0; text-align: center; "><img alt="" border="0" width="320" data-original-height="421" data-original-width="582" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhbVxP9ODFkPXtjjfi8o48W8IyUNZJivfe0oU_y-GEiU4Ak8QRUDLxzdsKuhpf8UmzNF0aSvPBCW9APV3u-iUeYYlP9pXPi0FbaK7pGGFp4_Fe_MNYRMsdxMWQOc7uiT12cMWtvWAgAXNunOaKjJLB9cjWG3xcL0_8jZsQsIatCuPWskbFy1vG4/s320/Screenshot%20from%202023-06-20%2019-56-18.png"/></a></div>
<div class="separator" style="clear: both;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEj9_gn13sDXEhHQHlkRbBZytCQ-fOkDJlsmdocC66prwt61zBQj1d1UEdo9TM9zLAfwV1HfEIs-wOchHnEyUmQtdJHjmyL-h4uDGZlPbd5e-FlnMOZB9jbJMjpaYZQPe0odEAyhvHtDdF8z_YBJR_vI24ads-EwXujwU5SAMtXvJ8nKc1wcaY2-/s645/Screenshot%20from%202023-06-20%2019-57-43.png" style="display: block; padding: 1em 0; text-align: center; "><img alt="" border="0" width="320" data-original-height="447" data-original-width="645" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEj9_gn13sDXEhHQHlkRbBZytCQ-fOkDJlsmdocC66prwt61zBQj1d1UEdo9TM9zLAfwV1HfEIs-wOchHnEyUmQtdJHjmyL-h4uDGZlPbd5e-FlnMOZB9jbJMjpaYZQPe0odEAyhvHtDdF8z_YBJR_vI24ads-EwXujwU5SAMtXvJ8nKc1wcaY2-/s320/Screenshot%20from%202023-06-20%2019-57-43.png"/></a></div>
<h3>Data Visualizations for Gig Workers</h3>
Our lab is also planning on making extensive use of data visualization in our research on the gig economy. One common and easy-to-understand form of data visualization is the word cloud, which displays words at a size corresponding to their frequency in a given text. One of the resources our team was using described how to create a word cloud using a Python library; however, as I was building ProductWords as a web plugin, I needed to find a way to do this with JavaScript.
I found a JavaScript library called D3 which is a general-purpose solution for creating visual representations of data to be displayed on web pages. Combined with an extension for D3 created by Jason Davies, I was able to create word clouds based on the descriptions of the Amazon products in the list.
<div class="separator" style="clear: both;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjUG0eJWGtstkryzo2O7ZUa6W4QwI_KEsz7G8hM7sHBOTRaC6LeLkF_utJ0ma3MmDpgTxDuQ_qTSARDtLMIc7RSQW7hTPZITW4rs1FCyHQh-b5J1KFVomEX9be5JOVcwUcKrgFBzpWjDrNwHfxTd38rJ2_PEh8oWo8oQNIGlpQ5a5SC3k7Co3qD/s611/Screenshot%20from%202023-06-20%2019-59-58.png" style="display: block; padding: 1em 0; text-align: center; "><img alt="" border="0" width="320" data-original-height="581" data-original-width="611" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjUG0eJWGtstkryzo2O7ZUa6W4QwI_KEsz7G8hM7sHBOTRaC6LeLkF_utJ0ma3MmDpgTxDuQ_qTSARDtLMIc7RSQW7hTPZITW4rs1FCyHQh-b5J1KFVomEX9be5JOVcwUcKrgFBzpWjDrNwHfxTd38rJ2_PEh8oWo8oQNIGlpQ5a5SC3k7Co3qD/s320/Screenshot%20from%202023-06-20%2019-59-58.png"/></a></div>
<div class="separator" style="clear: both;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgmFKmtwy7FlH6t4NoLuxFXHcIZrzgtJSpDoGYSaKLsd_ORBGvWYg8txETFzya5vUH0Cw2YCccd52uUC_4kY4ynUB7suHkx3yJqwLTE99k-KfaQO62E8KM_YIUO7Q6qfYSfQnBo0XuLrSHLiAE3FQRfzA5kj2H-Iz5saVQMGpc0urif25Mz90Cp/s730/Screenshot%20from%202023-06-20%2020-00-41.png" style="display: block; padding: 1em 0; text-align: center; "><img alt="" border="0" width="320" data-original-height="499" data-original-width="730" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgmFKmtwy7FlH6t4NoLuxFXHcIZrzgtJSpDoGYSaKLsd_ORBGvWYg8txETFzya5vUH0Cw2YCccd52uUC_4kY4ynUB7suHkx3yJqwLTE99k-KfaQO62E8KM_YIUO7Q6qfYSfQnBo0XuLrSHLiAE3FQRfzA5kj2H-Iz5saVQMGpc0urif25Mz90Cp/s320/Screenshot%20from%202023-06-20%2020-00-41.png"/></a></div>
<h3>Other Technical Aspects</h3>
One of the main reasons I decided to make this web plugin was that I wanted to practice some of the features of web plugins that we might want to use for our lab’s research. With ProductWords, I implemented web scraping (pulling the Amazon item description and price information), a popup page, and communication between the web-scraper background script and the popup script.<br>
ProductWords is not a very useful plugin yet, but I got some good practice implementing the features that will probably be necessary for any version of the web plugin(s) our lab will work on, and maybe it could even be used as a jumping-off point that gets evolved into our final product.
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-54375833669932658962023-01-27T19:30:00.013-07:002023-01-28T13:56:57.414-07:00ChatGPT: A Tool for Worker Collective Action?<div class="separator" style="clear: both;"><a href="https://external-content.duckduckgo.com/iu/?u=https%3A%2F%2Fcdn.greatlifepublishing.net%2Fwp-content%2Fuploads%2Fsites%2F4%2F2017%2F05%2F24153806%2Fwaves_feat-728x381.jpg&f=1&nofb=1&ipt=b8f96173823c5bddad1fff9bde1f87b4d4d7a01479339615d5f1c62a8314540b&ipo=images" style="display: block; padding: 1em 0; text-align: center; "><img alt="" border="0" width="320" data-original-height="280" data-original-width="500" src="https://external-content.duckduckgo.com/iu/?u=https%3A%2F%2Fcdn.greatlifepublishing.net%2Fwp-content%2Fuploads%2Fsites%2F4%2F2017%2F05%2F24153806%2Fwaves_feat-728x381.jpg&f=1&nofb=1&ipt=b8f96173823c5bddad1fff9bde1f87b4d4d7a01479339615d5f1c62a8314540b&ipo=images"/></a></div>
Are you tired of feeling powerless in the face of workplace injustices? Have you ever wondered how technology could help amplify your voice and bring about real change in the workplace? While some may fear <a href="https://chat.openai.com/chat">ChatGPT</a> as a <a href="https://www.theatlantic.com/ideas/archive/2023/01/chatgpt-ai-economy-automation-jobs/672767/">threat to job security</a>, we see it as a powerful tool for workers, especially for helping to bring change in the workplace.<br><br>
<h3>Challenges in Worker Collective Action.</h3>
<a href="https://www.history.com/topics/19th-century/labor">Worker collective action</a> is an essential tool for promoting fair treatment and rights in the workplace. By banding together, workers can make their voices heard and bring about real change. The rise of platforms for worker collective action, such as <a href="https://home.coworker.org/">Coworker.org</a>, has made it easier than ever for workers to share petitions, build support, and mobilize volunteers. However, participating in such actions can be challenging for workers due to the time and resources required. Traditional recruitment methods, such as door-to-door campaigns, can be <a href="https://journals.sagepub.com/doi/pdf/10.1177/1461444813518391">inefficient</a>, and social media or other digital platforms <a href="https://www.microsoft.com/en-us/research/publication/botivist-calling-volunteers-to-action-using-online-bots/">can still be time-consuming for activists to use</a>. That's where ChatGPT comes in.<br><br>
<h3>ChatGPT for Worker Activism.</h3>
As a large language model, ChatGPT can assist workers in <a href="https://journals.sagepub.com/doi/pdf/10.1177/1461444813518391">crafting compelling</a> and <a href="https://dl.acm.org/doi/pdf/10.1145/764008.763957">persuasive messages</a> to share on social media, inviting people to join their movement, and sharing their struggles. The complexity of social media strategies can make it difficult for activists to keep up with new technology and predict the outcomes of different strategies, but ChatGPT can help workers overcome these challenges by providing them with the knowledge and tools needed to effectively mobilize people. With ChatGPT, workers can reach a wider audience, influence behavior, and ultimately achieve their goals.
It can also help workers to research and find relevant information and resources, such as laws and regulations protecting their rights, as well as connect with relevant organizations and advocacy groups. Additionally, ChatGPT could be used to create educational materials, such as brochures, flyers, and infographics, that workers can use to educate others about their rights and the injustices they face.<br><br>
As technology continues to advance, it is improtant for workers to take advantage of these tools to make their voices heard and promote change. We believe ChatGPT can play a role in helping workers in this endeavor, making collective action more accessible and effective.<br><br>
**This post was partially edited with the help of ChatGTP :)
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-31328769172889873722022-12-11T11:00:00.003-07:002022-12-11T11:01:13.197-07:00Human Centered AI Live Stream: Sota Researcher!<div class="separator" style="clear: both;"><a href="https://i.imgur.com/NQIyrwq.jpeg" style="display: block; padding: 1em 0; text-align: center; "><img alt="" border="0" height="320" data-original-height="800" data-original-width="643" src="https://i.imgur.com/NQIyrwq.jpeg"/></a></div>
The research engineer Phil Butler from our lab is starting a new live stream on Human Centered AI. Through his live stream he will help you design and implement AI for people.<br>
<ul>
<li>In each live stream you will learn how to design and create AI for people from start to finish. He will teach you how to use different design methologies such as mockups,storyboards, service design; as well as different AI models and recent state of the art techniques. In each live stream you will have code that will help you to have a complete AI for people project. </li>
<li>Some of the topics he will cover in his live stream include: Understanding and Detecting Bias in AI; Design principles for Designing Fair and Just AI; How to Create Explainable AI.</li>
<li>The streams will benefit anyone who wants to learn how to create AI on their own, while also respecting human values. </li>
<li>The stream will help people to learn about how to implement AI using state of the art techniques (which is key for getting top industry jobs), while also being ethical and just about the AI that is created. </li>
</ul>
<br> <br> Join us! <a href="https://www.youtube.com/@sotasearcher">https://www.youtube.com/@sotasearcher</a>
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-71025418494683522512022-12-11T10:19:00.004-07:002022-12-11T10:20:26.155-07:00Designing Public Interest Tech to Fight DisinformationOur research lab organized a series of talks with NATO around how to design
public interest infrastructure to fight disinformation globally. Our
collaborator Victor Storchan wrote this great piece on the topic:<br />
Disinformation has increasingly become one of the most preeminent threats as
well as a global challenge for democracies and our modern societies. It is now
entering in a new era where the challenge is twofold1: it has become both a
socio-political problem and a cyber-security problem. Both aspects have to be
mitigated at a global level but require different types of responses.<br />
Let’s first give some historical perspective.<br />
<ul>
<li>Disinformation didn’t emerge with automation and social networks platforms of
our era. In the 1840s Balsac was already describing how praising or denigrating
reviews were spreading in Paris to promote or downgrade publishers of novels or
the owners of theaters. Though, innovation and AI in particular gave rise to
technological capabilities to threat actors that are now able to scale
misleading content creation.</li>
<li>More recently, in the 2000s the technologists
were excited about the ethos around moving fast and breaking things. People were
basically saying “let’s iterate fast, let’s shift quickly and let's think about
the consequences later.”</li>
<li> After the 2010s, and the rise of deep learning
increasingly used in the industry, we have seen a new tension emerging between
velocity and<a href="https://arxiv.org/abs/2209.09125" target="_blank"> validation</a>. It was not about the personal philosophy of the
different stakeholders asking for going “a little bit faster” or “a little bit
slower” but rather about the cultural and organizational contexts of most of the
organizations.</li>
<li> Now, AI is entering the new era of foundation models. With
large language models we have consumer-facing powered tools like search engines
or recommendation systems. With generative AI, we can turn audio or text in
video at scale very efficiently. The technology of foundation models is at the
same time becoming more accessible to users, cheaper and more powerful than
ever. It means better AI to achieve complex tasks, to solve math problems, to
address climate change. However, it also means cheap fake media generation
tools, cheap ways to propagate disinformation and to target victims.</li>
</ul> <br>
This is the moment where we are today. Crucially, disinformation is not only a
socio-political problem but also a cyber-security problem. Cheap deep-fake
technology is commoditized enabling targeted disinformation where people will
receive specific, personalized disinformation through different channels (online
platforms, targeted emails, phone). It will be more fine grained. It has already
started to affect their <a href="https://www.scopus.com/record/display.uri?eid=2-s2.0-84991489613&origin=inward&txGid=62c2fed975f9a74d7414daf7ab5c722c" target="_blank">lives</a>, their <a href="https://www.pnas.org/doi/pdf/10.1073/pnas.1320040111" target="_blank">emotions</a>, their <a href="https://www.financemagnates.com/cryptocurrency/how-did-deepfake-tech-drain-a-brazilian-crypto-exchange-out-of-liquidity">finances</a>, their health
etc.<br><br>
<b>The need for a multi-stakeholder approach as close as possible to the AI
system design.</b>The way we mitigate disinformation as a cyber-security problem is
tied to the way we are deploying large AI systems and to the way we evaluate them. We need new auditing tools and third parties <a href="https://arxiv.org/pdf/2112.07773.pdf">auditing procedures</a> to make sure that those deployed systems are trustworthy and robust to adverse threats or to toxic content dissemination. As such AI safety is not only an engineering problem but it is really a multi stakeholder challenge that will only be addressable if non-technical parties are included in the loop of how we design the technology. Engineers have to collaborate with experts in cognition, psychologists, linguists, lawyers, journalists, civil society in general). Let’s give a concrete example: mitigating disinformation as a cyber security problem means protecting the at-risk user and possibly curing the affected user. It may require access to personal and possibly private information to create effective counter arguments. As a consequence it implies arbitrating a tradeoff between privacy and disinformation mitigation that engineers alone cannot decide. We need a multi stakeholder framework to arbitrate such tradeoffs when building AI tooling as well as to improve transparency and reporting.<br><br><br>
<b>The need for a macroscopic multi-stakeholder approach.</b> Similarly, at a macroscopic level, there is a need for a profound global cooperation and coalition of researchers to address disinformation as a global issue. We need international cooperation at a very particular moment in our world which is being reorganized. We are living in a moment of very big paradox: we see new conflicts that emerge and structure the world and at the same time, disinformation requires international cooperation. At the macroscopic level, disinformation is not just a technological problem, it is just one additional layer on top of poverty, inequality, and ongoing strategic confrontation. Disinformation is one layer that adds to the international disorder and that amplifies the other ones. As such, we also need a multi stakeholder approach bringing together governments, corporates, universities, NGOs, the independent research community etc… Very concretely, Europe has taken legislative action DSA to regulate harmful content but it is now clear that regulation alone won’t be able to analyze,detect, and identify fake media. To that regard, the Christchurch call to action summit is a positive first step but did not lead yet to a systemic change.<br><br>
<b>The problem of communication.</b> However, the communication between engineers, AI scientists and non-technical stakeholders generates a lot of <a href="https://arxiv.org/pdf/2206.09511.pdf" target="_blank">friction</a>. Those multiple worlds don't speak the same language. Fighting disinformation is not only a problem of resources (access to data and access to compute power) but it is also a problem of communication where we need new processes and tooling to redefine the way we collaborate in alliance to fight disinformation. Those actors are collaborating in a world where it is becoming increasingly difficult to understand AI capabilities and as a consequence to put in place the right mechanisms to fight adverse threats like disinformation. It is more and more difficult to really analyze the improvement of AI. It is what Gary Marcus is calling the demoware effect: a technology that is good for demo but not in the real world. It is confusing people and not only political leaders but also engineers (Blake Lemoine at Google). Many leaders are assuming false capabilities about AI and struggle monitoring it. Let us give two reasons to try to find the causes of this statement. First, technology is more and more a geopolitical issue which does not encourage more transparency and more accountability. Second, information asymmetry between the private and public sectors and the gap between the reality of the technology deployed in industry and the perception of public decision-makers has grown considerably, at the risk of focusing the debate on technological chimeras that distract from the real societal problems posed by AI like disinformation and the ways to fight it.
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-78545921099607909202022-11-20T10:40:00.009-07:002022-12-04T09:34:07.718-07:00List of MIT Tech Review Inspiring Innovators!We are part of the amazing network of the 35 Innovators under 35 by the MIT Tech Review. We got invited to their EmTech Event and had amazing dinner with other innovators and people having an impact in the field. We are very thankful with Bryan Bryson for the invitation, and we also wanted to congratulate him and his team for all the work done to build such a vibrant innovation ecosystem.<br>
<br><br>
<br>
I share below a list of some of the innovators I meet. Keep an eye on them!<br><br>
*<b>Setor Zilevu (Meta and Virginia Tech</b>). Working at the intersection of human-computer interaction and machine learning to create semi-automated, in-home therapy for stroke patients. After his father suffered a stroke, Zilevu wanted to understand how to integrate those two fields in a way that would enable patients at home to get the same type of therapy, including high-quality feedback, that they might get in a hospital. The semi-automated human-computer interaction, which Zilevu calls the “tacit computable empower” method, can be applied to other domains both within and outside health care, he says.<br><br>
<b>Sarah B. Nelson</b> is Chief Design Officer and Distinguished Designer for Kyndryl Vital, Kyndryl’s designer-led co-creation experience. From the emergence of the web through the maturity of user experience practice, Sarah is known throughout the design industry as a thought leader in design-led organizational transformation, participatory, and forward-looking design capability development. At Kyndryl, she leads the design profession, partnering with technical strategists to integrate experience ecosystem thinking into the technical solutions. Sarah is an encaustic painter and passionate surfer.<br><br>
<b>*Moses Namara (Meta and Clemson University</b>). Namara co-created the Black in Artificial Intelligence graduate application mentoring program to help students applying to graduate school. The program, run through the resource group Black in AI, has mentored 400 applicants, 200 of whom have been accepted to competitive AI programs. It provides an array of resources: mentorship from current PhD students and professors, CV evaluations, and advice on where to apply. Namara now sees the mentorship system evolving to the next logical step: helping Black PhD and master’s students find that first job.<br><br>
*<b>Joanne Jang (OpenAI)</b>. Joanne Jang is the product lead of DALL·E, an AI system by OpenAI that creates original images and artwork from a natural language description. Joanne and her team were responsible for turning the DALL·E research into a tool people can use to extend their creative processes and for building safeguards to ensure the technology will be used responsibly. The DALL·E beta was introduced in July 2022 and now has more than 1 million users.<br><br>
<b>Daniel Salinas (Colombia)
</b>
Su ‘start-up’ monitoriza las plantas con nanotecnología al conectarlas con ordenadores y facilita la descarbonización. Los humanos tienen 'ceguera a las plantas'. Nuestros sesgos nos impiden percibirlas como sí hacemos con los animales. Esta desconexión planta-humano lleva a que los proyectos de plantar árboles para capturar carbono frente a la crisis climática no sean sostenibles si la reforestación no se mantiene en el tiempo. El estudiante de Emprendimiento colombiano Daniel Salinas descubrió la falta de infraestructuras en la lucha para la descarbonización con una 'start-up' de plantación de árboles. El joven recuerda: "Cada vez que íbamos al terreno teníamos problemas". Para romper esta desconexión entre personas y árboles, Salinas ha creado una interfaz planta-ordenador que permite hacer un seguimiento de la vegetación con su start-up Superplants. Con esta aportación, Salinas ha logrado ser uno de los Innovadores menores de 35 Latinoamérica 2022 de MIT Technology Review en español.<br>
<img src="https://media-exp1.licdn.com/dms/image/C4E22AQFdCaqDjb2Jzg/feedshare-shrink_1280/0/1668964936151?e=1671667200&v=beta&t=7dvjfarC_ZigC52CM1p4UGC0eiW3RcUb0MrNPeOkwv8" alt="Girl in a jacket" width="500" height="200";>
<br>
Relevant References:<br>
-https://www.building-up.org/knowledgehub/innovadores-menores-de-35-latinoamrica-2022<br>
-https://event.technologyreview.com/emtech-mit-2022/speakers<br>
-https://www.technologyreview.com/innovator/setor-zilevu/<br>Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-54432953390820092542022-11-11T17:42:00.006-07:002022-11-11T19:22:28.107-07:00Recap: AAAI Conference on Human Computation and Crowdsourcing (HCOMP 2022)<img src="https://pbs.twimg.com/media/FhUjbPiWYAUJD41?format=jpg&name=900x900" alt="Girl in a jacket" style="width:550px;height:500px;">
<br><br>This week we attended the AAAI conference on Human Computation and Crowdsourcing (HCOMP'22). We were excited about attending for several reasons: (1) we were organizing HCOMP's CrowdCamp, excited about having the power to drive the direction of this event within the conference!, (2) it was the 10-year anniversary of the conference and we were elated to reflect collectively on where we have come as a field over the years, (3) we chaired one of the keynotes of HCOMP, in particular, our PhD hero, Dr. Seth Cooper, and (4) we had an important announcement to share with the community!<br>
WE WILL BE GENERAL CO-CHAIRS OF HCOMP’23!<br><br>
<h2>Organizing CrowdCamp. </h2>
This year, Dr. Anhong Guo from the University of Michigan and me had the honor of organizing HCOMP's CrowdCamp, a very unique part of the HCOMP conference. It is a type of mini hackathon where you get together with crowdsourcing experts and define the novel research papers and prototypes that push forward the state of the art around crowdsourcing. Previous CrowdCamps led to key papers in the field, such as the Future of Crowd Work paper and my own CHI paper on Subcontracting Micro Work.<br><br>
This year, when we put out the call for CrowdCamp, we witnessed an interesting dynamic. A large number of participants were students, novices to crowdsourcing, but they had great interest in learning and then impacting the field. This dynamic reminded me of what I had encountered when I organized my first hackathon, FixIT: the participants had great visions and energy for changing the world! But they also had limited skills to execute their ideas. They lacked data to determine if their ideas were actually something worth pursuing. To address these challenges, in the past, I gave hackathon participants bootcamps to ramp up their technical skills (this allowed them to execute some of their visions). We also taught these participants about human centered design to empower them to create artifacts and solutions that match people's needs, and not a hammer in need of nails.<br><br>
For CrowdCamp, we decided to do a similar thing:<br>
We had a mini-bootcamp, organized by Toloka (a crowdsourcing platform), that explained how to design and create crowd-powered systems. The bootcamp started with a short introduction on what is crowdsourcing, common types of crowdsourcing projects (like image/text/audio/video classification) and interesting ones (like side-by-side comparison, data collection and spatial crowdsourcing). After that, the bootcamp introduced the Toloka platform and some of its unique features. Then the bootcamp briefly presented Toloka Python SDK (Toloka-Kit and Crowd-Kit) and moved to an example project on creating a crowd powered system, especially a face detection one. The code used in the Bootcamp is in the following Google Collab:
<a href="https://colab.research.google.com/drive/13xef9gG8T_HXd41scOo9en0wEZ8Kp1Sz?usp=sharing">https://colab.research.google.com/drive/13xef9gG8T_HXd41scOo9en0wEZ8Kp1Sz?usp=sharing</a>.
<br><br>
We taught human centered design, and had a panel with real world crowdworkers who shared their experiences and needs. The participants were empowered to design better crowdworkers and create more relevant technologies for them, as well as technologies that would better coordinate crowdworkers to produce higher quality work. The crowdworkers who participated in CrowdCamp all came from Africa, and they shared how crowd work had provided them with new job opportunities that were typically not available in their country. Crowd work helped to complement their expenses (a side job). They were motivated to participate in crowd work for the additional money received, also knowing that they were contributing to something bigger than themselves (e.g., labeling images that will ultimately help to power self-driving cars.) Some of the challenges these crowdworkers experienced included unpaid training sessions. It was unclear sometimes whether the training sessions were worth it or not. They also discussed the importance of building worker communities.<br><br>
CrowdCamp ended up being a success with over 70 people who had registered and then created a number of different useful tools for crowdworkers. The event was hybrid with people on the east coast joining us at Northeastern university. We had delicious pizza and given that we were in Boston, delicious Dunkin Donuts :)<br><br>
<h2>Chairing Professor Seth Cooper's Keynote.</h2>
We had the honor of chairing the keynote of Professor Seth Cooper, an Associate Professor at the Khoury College of Computer Sciences. He previously worked for Pixar Animation Studios and Electronic Arts, a big game maker. Seth is also the recipient of an NSF career grant.
Professor Cooper’s research has focused on using video games and crowdsourcing techniques to solve difficult scientific problems. He is the co-creator and lead designer, as well as developer of Foldit, a scientific discovery game that allows regular citizens to advance the field of biochemistry. Overall, his research combines scientific discovery games (particularly in computational structural biochemistry), serious games, and crowdsourcing games. A pioneer in the field of scientific discovery games, Dr. Cooper has shown video game players are able to outperform purely computational methods for certain types of structural biochemistry problems, effectively codifying their strategies, and integrating them in the lab to help design real synthetic molecules. He has also developed techniques to adapt the difficulty of tasks to individual game players and generate game levels.<br><br>
Seth’s talk discussed how he is using crowdsourcing to improve video games, and video games to improve crowdsourcing. What does this mean? In his research, Professor Cooper integrates crowd workers to help designers improve their video games. For example, he integrates crowds to help them test just how hard or easy the game they are creating is. It enables designers to identify how easy it is for gamers to advance within the different stages of a game. The integration of crowdworkers allows gamers to easily iterate and improve their video game. Dr. Cooper is also integrating gaming to improve crowdsourcing. In particular, he has studied how he can integrate games to improve the quality of work produced by crowd workers.<br><br>
During the Questions of Professor Cooper, some interesting questions emerged:<br>
What types of biases do crowdworkers bring to the table when co-designing video games? It was unclear whether crowdworkers are actually similar to how typical gamers would play a video game. Hence, the audience wondered just how much designers actually use the results of the way crowdworkers engage with a video game. Professor Seth mentioned that in his research, he found that crowdworkers are similar to typical gamers in playing games. A difference is that typical gamers (voluntarily playing a game instead of getting paid to play) will usually focus more on the aspects of the game they like the most. Crowdworkers will focus on exploring the whole game instead of focusing on particular parts (because of the role that the payments play). Perhaps, these crowdworkers feel that by exploring the whole game, they are better showcasing to the requester (designer) that they are indeed playing the game and not slacking off. Some people have a gaming style that focuses on the "catch-them all'' approach (an exploratory mode). However, the "catch-them-all" term is used in reference to Pokemon, where people are interested in being able to explore the entire game and collect all the different elements (e.g., Pokemons).<br><br>
How might we integrate game design to help crowdworkers learn? Dr. Flores-Saviaga posed an interesting question about the role games could play in facilitating the career development of these workers. Professor Seth expressed an interest in this area while also mentioning that you can imagine that workers instead of earning badges within the game could earn real certificates that translate into new job opportunities.<br><br>
What gave him confidence that the gaming approach in crowdsourcing is worth pursuing? When Foldit came out, it was unclear that gaming would actually be useful for mobilizing citizen crowds to complete complex scientific tasks. The audience wanted to know what led him to explore this path. Professor Cooper explained that part of it was taking a risk down a path he was passionate about: gaming. I think for PhD students and other new researchers starting out, it can be important to trust your intuition and conduct research that personally interests you. In research, you will take risks, which makes it all the more exciting :)<br><br>
<img src="https://pbs.twimg.com/media/FhUrFfhWQAAGte_?format=jpg&name=medium" alt="Girl in a jacket" style="width:400px;height:400px;">
<h2>Dr. Jenn Wortman’s Keynote.</h2>
We greatly enjoyed the amazing keynote given by Dr. Jenn Wortman Vaughan (@jennwvaughan) at HCOMP 2022. She presented her research in Responsible AI, specially interpretability and fairness in AI systems.<br><br>
A takeaway is that there are challenges in the design of interpretability tools for data scientists such as IntrepretML or SHAP Python package, where they found that these tools lead to over-trusting and misusing how ML models work. For more info, look at her CHI 2020 paper: "Interpreting Interpretability”<br><br>
<h2>Dr. Jeffrey Bigham’s Keynote.</h2>
An incredible keynote given by Dr. Jeffrey Bigham at HCOMP 2022. He presented work developed in Image Description for 17 years! He showed different connections (loops) in finding the right problem and right solution in Image Description, such as Computer Vision, Real-Time Recruitment, Gig Workers, Conversations with the Crowd, Datasets, etc.<br><br>
A takeaway is that there could be different interactions or loops in the process of Applying Machine Learning and HCI as seen in the image below from the Problem selection until the Deployment of the system.<br><br>
<h2>Doctoral Consortium.</h2>
The HCOMP doctoral consortium was led by Dr. Chien-Ju Ho and Dr. Alex Williams. The consortium is an opportunity for PhD students to share their research with crowdsourcing and human computation experts. Students have the opportunity to meet other PhD students, industry experts, and researchers to expand their network and receive mentoring from both industry and academia. Our lab participated in the proposal “Organizing Crowds to Detect Manipulative Content.” A lab member, Claudia Flores-Saviaga, presented the research she has done in this space for her PhD thesis.<br><br>
<h2>Exciting news for the HCOMP community!</h2>
The big news I want to share is that I have the honor of being the co-organizer of next year's HCOMP! I will co-organize it with Alessandro Bozzon and Michael Bernstein. We are going to host the conference in Europe, and it will be united with the Collective Intelligence conference. Our theme is about reuniting and helping HCOMP grow in size by connecting with other fields, such as human centered design, citizen science, data visualizations, and serious games. I am excited to have the honor and opportunity to build the HCOMP conference.
<img src="https://pbs.twimg.com/media/FhRPFugXgAA08KU?format=png&name=900x900" alt="Girl in a jacket" style="width:400px;height:300px;">
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-13801349172782600972022-06-08T22:20:00.053-06:002022-06-09T04:35:19.146-06:00Summer: Work in the Age of Intelligent Machines<div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiDfzKz4OGrUC6VAzd_0Vft6xsWNpiydgb8hoZXbxHrdhsqLYmmXjWsBgwaC9fNiTcWuPAqXGrJGlZSrXnzbreW4RQTuFtSLvacghWQ-7bHCdP7KXpRUEAl15EId24R7NPEWkggtBe1G9_DhYzWb7wDWwmEZo6TO3Hewe7Ujkh7kH_iz1OcDQ/s921/banner%20civic%20AI%20lab.jpeg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img alt="photos of dr. savage at the conference" border="0" data-original-height="218" data-original-width="921" height="128" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiDfzKz4OGrUC6VAzd_0Vft6xsWNpiydgb8hoZXbxHrdhsqLYmmXjWsBgwaC9fNiTcWuPAqXGrJGlZSrXnzbreW4RQTuFtSLvacghWQ-7bHCdP7KXpRUEAl15EId24R7NPEWkggtBe1G9_DhYzWb7wDWwmEZo6TO3Hewe7Ujkh7kH_iz1OcDQ/w537-h128/banner%20civic%20AI%20lab.jpeg" title="Photos of the event" width="537" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><br /></div><div>I had the honor to start the summer attending and giving a keynote at the Work in the Age of Intelligent Machines (WAIM) Summer Conference. The conference, which is partially funded by the US National Science Foundation aims to create a network of researchers that come together on the topic of work in the age of intelligent machines. In this post I will share a little bit of the discussions and activities we had at WAIM. </div><br /><br /><h4 style="text-align: left;"><span style="font-size: medium;">Opening Remarks.</span></h4><div><div>The conference started with opening comments from the conference organizers, whose profiles I share below. Note that I share their profiles because I find interesting how diverse the organizers were (which I think was key for organizing this type of conference):</div><br /><b>Dr. Kevin Crowston.</b> Distinguished Professor of Information Science at Syracuse University. He received his Ph.D. in Information Technologies from the Massachusetts Institute of Technology (MIT). His research examines new ways of organizing by the use of information technology. He approaches this issue in several ways: empirical studies of coordination-intensive processes in human organizations (especially virtual organization); theoretical characterizations of coordination problems and alternative methods for managing them; and design and empirical evaluation of systems to support people working together.<br /><b>Dr. Jeffrey Nickerson.</b> Professor and Associate Dean of research in the School of Business at Stevens Institute of Technology. His research focuses on different aspects of collective creativity, in particular the way crowds and communities design digital artifacts: 3D printing designs, systems designs, source code, and articles. He has a Ph.D. in Computer Science.<br /><b>Dr. Ingrid Erickson.</b> Associate Professor of Information Science at Syracuse University. Ph.D. from the Center for Work, Technology and Organization in the Department of Management Science and Engineering at Stanford University. She is a scholar of work and technology, currently fascinated by the way that mobile devices and ubiquitous digital infrastructures are influencing how we communicate with one another, navigate and inhabit spaces, and engage in new types of socio-technical practices.</div><div><br /></div><div>
The opening remarks began with an overview of what WAIM had become, how the network had grown, and impact obtained through time. The main goals of WAIM was to create a network of researchers that push forward together investigations on work and machines. For instance, what role do machines have in labor? How do they change the work dynamics? What type of new science fiction realities do we create from integrating intelligent machines in the workplace? What type of futures around work and machines do we want to avoid? What new power dynamics emerge from integrating machines in the workplace? Personally, WAIM has become a key space to build key research collaborations. From prior WAIM conferences, I was able to start working with Professor Jarrahi from UNC, Professor Matthew Lease from UT Austin, Professor Steve Sawyer from Sycracruse University, and Professor Michael Dunn from the University of Albany. Overall WAIM became a great place to connect with academics in the United States who also had an interest in the future of work, and wanted to push forward a new future of work reality with machines. I have particularly liked that I have been able to connect with academics who are not just in computer science, but also in economy, information schools, business schools. The diversity has certainly helped bring new perspectives to my research in the space. It was inspiring and energizing way to start the conference with excellent opening statements and remembering all that we had been able to start building together around work and intelligent machines.</div><div><br /></div><div><h4 style="text-align: left;"><span style="font-size: medium;">Opening Keynote: Saiph Savage.</span></h4>
I then gave the opening keynote. You can access my slides here: XX. Overall, I discussed how to design meta systems that can empower gig workers to be able to design and create their own tools. We had very interesting discussions around how to motivate workers to collaborate with each other to create their own tools (collaborations are hard because gig platforms also promote competition between workers). I also discussed how to facilitate quality data sharing between workers, and how might you engage the other stakeholders to participate in the design process (I use value sensitive design in this front).<br /><br /><h4 style="text-align: left;"><span style="font-size: medium;">Panel: "Illuminating the Human-Technology Frontier" with Bledi Taska (Emsi Burning Glass), Nick Diakopoulus (Northwestern), Sarah Leibovitz (University of Virginia).</span></h4></div><div><br />
Afterwards, we had a panel on "Illuminating the Human-Technology Frontier", with participation from Bledi Taska (Emsi Burning Glass), Nick Diakopoulus (Northwestern), Sarah Leibovitz (University of Virginia), and the moderator was Jeffrey Nickerson. Dr. Bledi works at Emsi Burning Glass which provides the nation’s premier labor market data. Dr. Bledi discussed about how jobs are changing. From the data in his company he has observed that most changes occurs in jobs that involve technology, and what changes the most are the type of key skills that these jobs need (i.e., top 40 skills needed for the job). Professor Diakopoulus then discussed the role that AI plays in journalism and argued for the importance of better understanding the values that exist in each profession, to define how data looks like, what defaults are chosen, what inputs the system has. It is important to understand the tensions that can exist when defining technology for journalists, especially based on their values. For example, you have to think about: Is there sufficient transparency in the AI so that the journalist feels comfortable with the results? Overall, Prof. Diakopoulus argued for the importance of thinking about the intelligent systems and how to design them to match the values of a profession. He also argued for the importance of thinking about the edge that humans have over AI. Journalists have to negotiate with sources to get their information out of them. That is something that AI still cannot do. Additionally, the journalists need to have creativity in how they communicate and present their story to effectively engage their audience. So, it is important to think about what is the edge that humans have, how might we design technology that does not focus on replacing journalists, but instead, enhances their work. Within this setting of intelligent systems it is also important to think about how intelligent systems have also made the job of journalists more difficult. Particularly, e.g., we have now a number of bots that are contaminating the information ecosystem and journalists have to learn how to navigate and cut through the automated systems that might be spreading political lies. </div><div><br /></div><div><br /></div><h4 style="text-align: left;"><span style="font-size: medium;">Panel: Future of Work at the Human-Technology Frontier, <br />with NSF Program Director Andruid Kerne.</span></h4><div>After the panel we had a fantastic lunch where conversations around the future of work with intelligent machines were common. We then heard insights Federal Government Employee, Dr. Kerne, who currently works at the US National Science Foundation (NSF), and is a Program Director for NSF's Future of Work at the Human-Technology Frontier program. He discussed what makes good research proposals for this program. He argued that good research proposals are proposals that include multiple views: the perspectives of workers, the insights about the new technologies that will be developed, and how it ultimately affects work. The program directors of this program also considered that if your research could go into another program by NSF, then you should just send it there. This highlight the importance of identifying what research is indeed very unique to the Future of Work at the Human-Technology Frontier. Dr. Kerne also discussed that for this program it was also important for the intellectual merit to involve multiple fields. The program has a commitment to look for proposals that bring together different fields, and that also function across fields, not just the silos of fields. He argued that research across different fields is necessary to address issues around the future of work. </div><div><br /></div><div>I also found particularly interesting that Dr. Kerne argued that for the program it was important to also think about and consider the negative broad impacts that can emerge in the research that is proposed. For instance, how might the technologies that researchers want to study, also facilitate the surveillance of workers, the deterioration of their work conditions, and also how might it result in the digital privacy violations of workers. </div><div><br /></div><div>Finally, Dr. Kerne presented examples of successful projects that had been funded. The projects were interesting precisely because they combined multiple fields. For example, there were projects around empowering the labor and entrepreneurship of indigenous communities working in computational ceramics. There were also research projects on Occupational Exoskeletons that involved the areas of Mechanical engineering, sociology, and economics. Other funded projects were on the Future of Automation in the Hospitality Industry (which also included academics in healthcare, HCI, and design). Overall, I found the talk from Dr. Kerne useful to better understand the things NSF's program on the future of work might value to further ensure my success when I apply to future grants :)</div><div><br /></div><div><h4><span style="font-size: medium;">Panel: WAIM Fellows Presentations.</span></h4></div><div>A neat thing from this conference was that they were able to also have fellows! The conference funded the year long research of PhD students who conducted investigations on "Work in the Age of Intelligent Machines" ! The fellows included PhD students from UT Austin, Carnegie Mellon University, Georgia Tech, among others. It was also inspiring to see that all the PhD students who were funded were women! Their research included:</div><div><ul style="text-align: left;"><li><b>Investigating how AI Can be Integrated in Hiring Decisions (UT Austin). </b>Here the research argued that when people are working with AI based tools, they are not always experts, and they can have very different backgrounds and experiences with AI, which can influence how they use the technology. This research focused on studying how people's background in AI impacts how they use AI, as well as how they use the information that AI provides to them within the hiring process. <br /><br />The research first focused on understanding people's AI literacy. The work focused first on being able to measure people's literacy in AI and then studying how the literacy that was detected impacted how these individuals made decisions around hiring with AI. Some challenges within this research is: how do you measure AI literacy? In this particular work, they ended up using a taxonomy that already exists, that focuses on measuring how much people know certain AI concepts and how much they are able to create new AI applications. The research also found that it was important to study what people's understanding of AI is. People had very different perspectives about what AI was and what AI could do. Such understanding is important as it affects how much individuals trust and use the results that an AI based hiring system outputs to them. Some of the things the research focused on measuring around people's understanding of AI, was how much people knew that there were humans involved in the process of developing algorithms? Do they have a good idea of how algorithms work?
What do people think of AI integrated in the decision process? Do they trust it?
Do people have positive/negative opinions of AI? Do they trust it or just don’t like it? Currently the research fellow discussed how she is planning on running a national sample to make better sense of people's opinions and knowledge around AI.
She will then run an experiment where people run a Job evaluation task with AI to study how people's knowledge and perceptions around AI affects their behavior with the AI based job evaluation results. Based on her findings she plans to create also educational material on how AI can be best used for hiring given people's different backgrounds.<br /><br /><br /></li><li><b>Futures of AI Based Care Work (Georgia Tech). </b> The research studied how AI based technologies could be integrated for rural nurses to help reduce harms, and help them in their jobs. The research is also looking at the new labor that organizations have to take on when new technology comes into the picture (new types of indivisible labor!)<br /><br /></li><li><b>Peer Tools for Gig Workers (CMU). </b>The research discussed how gig workers are on their own to build their brand and figure out the work they have to do. The research of Yasmine, the Phd student leading the work, focuses on designing peer support systems for gig workers. For instance, she designed "Hire Peer", an interface design that offers new formative feedback on creative entrepreneurship so gig workers can help each other to grow and develop themselves as entreponeurs as well. Her research also focuses on helping workers to better brand themselves, and create an identity for themselves. Her research is heavily based on human centered design and participatory design. Given that her research is also interested in the creation of peer support networks of gig workers (i.e., communities) she has been also studying how can we design solutions that are well integrated into communities. She questions: How do we sustain community driven designs? How do we support community based research, even when there are institutional barriers that can hinder the analysis and reaching the communities? <br /><br /></li></ul>Overall, it was inspiring to see how the WAIM conference was able to fund and push forward the research of new top researchers in the future of work. I liked learning about the new research directions these scholars were exploring. I think to know where the field will go it is important and critical to listen to the new researchers. </div><div><br /></div><div><br /></div><div><h4 style="text-align: left;"><span style="font-size: medium;">Panel: </span>The Future Work of the Future of Work: MC Binz-Scharf (CUNY), Katie Pine (Arizona State University), Joel Chan (University of Maryland), <br />Moderator Ingrid Erickson</h4></div><div>This panel involved some of the people I admire the most! So it was a trill and pleasure to hear each panelist and the moderator. Dr. MC Binz-Scharf discussed how the future of work is bright. But not for everyone. It can depend on the privilege of the individual. She discussed how there are matrixes of oppression: we never have just one identity and together our different identities can result in different types of discriminations and harms that we can push onto others. It is thus important to think about what are the consequences that exist from the privilege of certain individuals? For example, in healthcare many times men get the privilege of being studied fully. While women are sometimes considered to be simply "small men". That privilege that men hold bring several misconceptions around best practices for treating women in the healthcare system. There are currently no standard of care for women. Women are thus likely to be misdiagnosed. Black women are 3 times more likely to die in child birth.
Inequality is codified in society and in the workplace. There are a number of structures that permit inequality. What type of jobs do women vs men get? The flexibility of gig work and certain types of jobs benefits some but not everyone. It is important to understand the socio-economic differences that can emerge from privilege. Technology also permeates inequality. Lots of related research, such as the books: Algorithms of Oppression; In Big Data We Trust. People trust the big data algorithms; but the programmer, the coders influence the design of the algorithms and the biases that exist. People trust algorithms without questioning these biases. MC also discussed how "DEI" is a new buzz word that has substitute "work inequality." However, it has also been different than affirming action. We need to understand the changes these new dynamics generate. MC argued for the following research directions: Studying the impact of technology on behavioral changes and diversity. She argues it is especially important to measure the effect of the DEI initiatives. An interesting point that MC made was that the people who enjoy privilege should take part in DEI initiatives. If you enjoy some privilege join the party and be part of DEI initiatives. She argues their integration is important as these individuals are who have the current power to push forward change. </div><div><br /></div><div>Similarly, Katie Pine (Arizona State University) and Joel Chang (University of Maryland) discussed about who gets to design the future of work.
People in the neighborhood want to participate in the design of the systems that control the neighborhood. But the problem is: Who actually gets to participate in the design? Workers are many times not able to design their own tools. It is a hard problem. How might we empower workers to design? How do we help workers to be proactive, not reactive in the designs they propose? How do we get there? Katie and Joel argued for the importance of creating coalitions to address the problem, as well as conducting Participatory Design. However, integrating participatory design is NOT as straightforward, especially because Participatory Design requieres a lot of work to really involve people in the design process. It can take a lot of work for the true needs to come out. It is also hard to know where should the design of tools by participants happen? You can bring people into the lab to design. But who gets to make it to the lab? (Many times it is elites who can travel to the laboratory because they might have more free time, or means of getting to the lab). Additionally, you could argue that researchers should just go to where the people are and do the participatory design where the individuals are. The problem with that approach is that there are a number of different barriers to go to their space, but then once you even arrive you have new issues that you also need to learn how to handle, such as power dynamics. The question then is: How do we create a shared space where people can all co-design? How do we design that space so people can really design together? There is also value in studying how much the designs that were created by gig workers were of any good. How do you evaluate them? No clear sense of quality. Is it time for a Meta-design for the future of work?</div><div><br /></div><h4 style="text-align: left;">Keynote 2 Youngjin Yoo (Case Western)</h4><div>The second keynote was with famous Professor Youngjin Yoo. He is the Elizabeth M. and William C. Treuhaft Professor in Entrepreneurship and Professor of Information Systems at the department of Design & Innovation at the Weatherhead School of Management, Case Western Reserve University. An Association of Information Systems Fellow, he is also WBS Distinguished Research Environment Professor at Warwick Business School, UK. and a Visiting Professor at the London School of Economics, UK. He is the founding faculty director of xLab at Case Western Reserve University. He has worked as Innovation Architect at the University Hospitals in Cleveland, overseeing the digital transformation efforts at one of the largest teaching hospital systems in the country. Before he returns to Case Western Reserve University, he was the Harry A. Cochran Professor of Management Information Systems and the Founding Director of Center for Design+Innovation at the Fox School of Business, Temple University where he was also the founder and Principal Investigator of Urban Apps & Maps Studios, an interdisciplinary initiative for digital urban entrepreneurship in Philadelphia. Previously, he was the Lewis-Progressive Chair of Management at Case Western Reserve University. He has taught digital innovation strategy at Indian School of Business, Aalto University in Finland, and Korean Advanced Institute of Science and Technology. He was a summer research fellow at NASA in summer of 2001 and spent a year as a research associate in 2003 – 2004 at NASA Glenn Research Center. He was also a visiting professor at Chalmers University of Technology in Sweden, Viktoria Institute in Sweden, Hitotsubashi University in Japan, Hong Kong City University, Yonsei University, Korea and Tokyo University of Science, Japan.
He holds a PhD in Information Systems from the University of Maryland. His research interests include: digital innovation and entrepreneurship, organizational genetics, societal use of technology and design. He has received over $4.5 million in research grant from National Science Foundation, NASA, James S. and John L. Knight Foundation, the Department of Commerce, National Research Foundation of Korea, and Samsung Electronics. His work was published at leading academic journals such as MIS Quarterly, Information Systems Research, Organization Science, the Communications of the ACM, and the Academy of Management Journal among others. He is Senior Editor of MIS Quarterly, the Journal of AIS, and the Journal Information Technology, and is on the editorial board of Organization Science, Scandinavian Journal of Information Systems, and Information and Organization. He was a former senior editor of the Journal of Strategic Information Systems and an associate editor of Information Systems Research and Management Science. He has worked with leading companies including Samsung Electronics, Samsung Economic Research Institute, American Greetings, Progressive Insurance, Goodyear Tire, Sotera Health, Bendix, Moen, Intel, Ford Motor Company, Andersen Consulting, IDEO, Gehry and Partners, University Hospitals in Cleveland, American Management Systems, Lotus, NASA, Parker Hannifin, Poly One and the Department of Housing and Urban Development. </div><div><br /></div><div>These were some of the main activities of the first day. The conference offered amazing breakfast, lunch, and reception events. These social gathering events during the conference were really nice for connecting and networking with other researchers (We also had very nice night walks to the white house! See picture above.) The next day we had in depth discussions on our papers and research proposals. I found this piece of the conference particularly useful for my own work, as I was able to better craft my research contribution in a much better way. I felt I was receiving coaching from top Olympic athletes that were guiding me to also be successful. Overall, it was a fantastic research event. Very useful for advancing on my papers, proposals, and tools I co-design and create with gig workers. It also opened new collaborations, and was overall great for pushing forward the network of researchers conducting investigations on Work in the Age of Intelligent Machines :)</div>Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-27181203640489241312020-08-11T10:36:00.027-06:002020-08-12T10:49:37.089-06:00Designing for the Future of Work During Covid19<iframe allow="autoplay; fullscreen" allowfullscreen="" frameborder="0" height="360" src="https://player.vimeo.com/video/446926681" width="500"></iframe>
<p><a href="https://vimeo.com/446926681">Solidarity and A.I. for Transitioning to Crowd Work during COVID-19</a> from <a href="https://vimeo.com/user121222317">Saiph Savage</a> on <a href="https://vimeo.com">Vimeo</a>.</p>
<p><span style="font-family: arial; font-size: medium;"></span></p><div class="separator" style="clear: both; text-align: center;"><span style="font-family: arial; font-size: medium;"><br /></span></div><span style="font-family: arial; font-size: medium;"> Last week, my research was featured in <a href="https://www.microsoft.com/en-us/research/theme/future-of-work/">Microsoft's New Future of Work Symposium</a>. I also helped chair a few sessions. The goal of the symposium was <span face="">to provide an open forum to explore:</span></span><span style="border: 0px none; box-sizing: inherit; font-size: medium; margin: 0px; padding: 0px; vertical-align: baseline;"><span style="font-family: arial;"> where the future of work should go towards, especially given the changes the COVID-19 pandemic has brought.</span></span><p></p><div><span style="font-family: arial; font-size: medium;"><br /></span></div><div><span style="font-family: arial; font-size: medium;"> Below I present a brief summary of some of the sessions I attended in the symposium: </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><h2 style="text-align: left;"><span style="font-family: arial; font-size: x-large;">🌟 Highlighted Talks: Gig Work, VR, & Disabilities. </span><span style="font-size: medium;"><br /></span></h2><span style="font-family: arial; font-size: medium;">My research was featured in this set of highlighted talks. <a href="https://www.microsoft.com/en-us/research/people/serintel/">Sean Rintel</a> </span><span style="font-family: arial; font-size: medium;">from Microsoft was our chair. </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;">Some of the talks included:</span></div><div><span style="font-family: arial; font-size: medium;"><br /><span style="color: #666666;"><b>Joshua McVeigh-Schultz & Katherine Isbister, SFSU and UC Santa Cruz, <span style="border: 0px none; box-sizing: inherit; margin: 0px; padding: 0px; vertical-align: baseline;">VR in Workplace Meetings: Learning from Social VR in ‘The Wild’</span></b></span></span></div><div><span style="font-family: arial; font-size: medium;"><b><br /></b></span><span style="font-family: arial; font-size: medium;">This talk discussed the ways in which VR could be used to improve work environments. I found particularly interesting the notion of how surreal animals (such as gigantic cats, similar to the ones in Alice in Wonderland) could be integrated into the work place to provoke conversations and interactions that might not be possible with traditional labor. This research in general is exploring how VR can create labor interactions we might not be able to have under traditional settings. </span><span style="font-family: arial; font-size: medium;"><span style="font-family: arial;">However,
here it is also important to think about how do we avoid turning the
workspace into an amusement park . How might we still maintain the
seriousness and professionalism that some employers and workers might
value? Other related questions that emerged for this work, was about how integrating VR might create a large cognitive load on workers, especially when they had to switch contexts (e.g., they are working from home and engaging in a virtual environment and meeting with their boss, but now they have to switch contexts and check on their toddler that exists just in the real world.) It is important to also consider how these technologies might make it easier for workers to switch between the different activities they do in their labor, but also in their personal lives. <br /><br /></span>This research in general falls under the areas of <a href="https://en.wikipedia.org/wiki/Serious_game">serious games</a>. My research lab has also been exploring <a href="https://dl.acm.org/doi/pdf/10.1145/3322276.3322359">storytelling and games in crowd work</a>. Especially how we could integrate play as a tangent activity into online labor. Below some examples of our play in crowd work research:<br /></span></div><div><br /><span style="font-family: arial; font-size: medium;"><img class="key-image" src="https://dl.acm.org/cms/attachment/49892038-7af8-4c1d-b583-bf5ecb5b7de3/3322276.3322359.key.jpg" /></span><br /><span style="font-family: arial; font-size: medium;"><img alt="Figure 2. A storyboard discussion prompt for speed-dating." data-selenium-selector="figure-image" height="343" src="https://d3i71xaburhd42.cloudfront.net/38d92a2deca949fd965313a1b9dda293056da9d9/4-Figure2-1.png" width="428" /><span style="color: #666666;"><b> </b></span></span></div><div><span style="font-family: arial; font-size: medium;"><span style="color: #666666;"><b><img alt="Figure 1. Turker Tales interface: Top panel shows a Turker-created scenario another Turker could see during a categorization HIT. Lower panel shows the message and GIF they see upon completing the HIT." data-selenium-selector="figure-image" height="540" src="https://d3i71xaburhd42.cloudfront.net/38d92a2deca949fd965313a1b9dda293056da9d9/1-Figure1-1.png" width="426" /> <br /></b></span></span></div><div><span style="font-family: arial; font-size: medium;"><span style="color: #666666;"><b> </b></span></span></div><div><span style="font-family: arial; font-size: medium;"><span style="color: #666666;"><b> </b></span></span></div><div><span style="font-family: arial; font-size: medium;"><span style="color: #666666;"><b>John Tang, Microsoft, </b></span><span style="border: 0px none; box-sizing: inherit; margin: 0px; padding: 0px; vertical-align: baseline;"><b><span style="color: #666666;">Early Indicators of the Effect of the Global Shift to Remote Work on People with Disabilities.</span><br /></b></span><b><br /></b>This research from Microsoft was exploring the ways that the pandemic was affecting the work of people with disabilities. One of the things I found interesting from this talk was that the work found that for people with disabilities one of the biggest challenges they faced was having to put on their cameras during meetings. Having to turn on their cameras was hard because many times this meant that the person with disabilities would have to "perform" for the people on the call (e.g., ensure they keep a certain posture in front of the camera, look a certain way), in addition to paying attention to what each person said and contributed (which was more difficult in a mediated communication settings, because the audio could be more noisy than meeting in person, or they might not have as clear an image of the people speaking to make out what they were saying) Additionally, many times people with disabilities had interpreters who were the ones who ended up interacting with their co-workers for them. Given that algorithms sometimes highlight the people who participate and interact with others, the people with disabilities felt left out of the conversation, erased (as the algorithms would never showcase them within the chat and video interfaces). I found this talk thought provoking and it helped me to start to think about how the technology we have and norms we establish (such as forcing everyone in a meeting to turn on their cameras), can create power struggles, especially put people with disabilities in a position where they are erased from conversations, and also are given additional cognitive loads.</span></div><div><span style="font-family: arial; font-size: medium;"><br /></span></div><div><span style="font-family: arial; font-size: medium;"><span style="color: #666666;"><b>Saiph Savage & Mohammad Jarrahi, Microsoft and UNC Chapel Hill. <span style="border: 0px none; box-sizing: inherit; margin: 0px; padding: 0px; vertical-align: baseline;">Solidarity and A.I. for Transitioning to Crowd Work during COVID-19</span> </b></span><br />Our research focused on combining worker solidarity and A.I. to design the future of work after COVID-19. Our video isshown on top of this post: <br /><br /><i>Due to the COVID-19 pandemic, a number of gig workers who engaged in location-based gig work (e.g., Taskrabbit, Care.com, or Wag) have had to transition to new jobs that are independent of location (e.g., online freelancing or crowd work). However, this has been a difficult transition. Especially because in this new environment, gig workers now have to compete globally for work, and they also have to focus on work interactions that are primarily online (instead of gig work that takes place within specific physical locations or within in-person meetings). In this paper, we build on our extensive research on gig work, gig literacy and the design of crowdsourcing systems, to present an intelligent architecture for helping workers transition to new gig jobs in times of global crisis. Our intelligent architecture uses machine learning and draws on collective action theory to introduce ``Solidarity Brokers.'' Our Solidarity Brokers are computational mechanisms that identify the best ways to build solidarity between workers with the purpose of mobilizing workers to help each other transition to new jobs. We finish by presenting a brief research agenda for intelligent tools that facilitate work transitions during the global pandemic and beyond.</i></span></div><div><span style="font-family: arial; font-size: medium;"><i><br /></i></span></div><div><p style="text-align: left;"><span style="font-family: arial; font-size: medium;">Next, I co-chaired with <a href="https://ischool.syr.edu/steven-sawyer/">Professor </a></span><span style="font-family: arial; font-size: medium;"><span face="" style="-webkit-text-stroke-width: 0px; background-color: rgba(255, 255, 255, 0.9); color: #202124; display: inline; float: none; font-family: "roboto", "robotodraft", "helvetica", "arial", sans-serif; font-style: normal; font-variant-caps: normal; font-variant-ligatures: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-decoration-color: initial; text-decoration-style: initial; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px;"><a href="https://ischool.syr.edu/steven-sawyer/">Steven B Sawyer</a> </span>the break out session of: </span></p><h2 style="text-align: left;"><span style="font-family: arial; font-size: x-large;"><b>✨"Employment: Hiring, onboarding, management, freelancing, on-demand, crowdwork, gig work" </b></span></h2></div><div><span style="font-family: arial; font-size: medium;">Some interesting things that were discussed here was that:</span></div><div><ul style="text-align: left;"><ul><li><span style="font-size: medium;">Instcart during the pandemic has been pushing a narrative of "The gig workers on this platform are heroes!" However, not all Instacart workers are buying this narrative of seeing themselves as heroes. They are bothered that the company is not doing more to keep workers safe. The narrative feels manipulative, especially when the workers are out risking their health just so somebody else can eat an ice-cream. Workers feel they are not getting essential food, but rather "buying the indulgences of an elite". <br /><br /></span></li></ul><ul><ul><li><span style="font-size: medium;">Interesting to learn how gig workers are identifying what tasks are "essential" and which ones are more "indulgence". <br /><br /></span></li><li><span style="font-size: medium;">It is important to consider how these hero narratives will change over time and whether they will have any long-lasting effects on how we view these gig workers.<br /><br /></span></li></ul><li><span style="font-size: medium;"> Instacart workers tend to be more white women. I found this interesting given that a large portion of <a href="https://techcrunch.com/2020/05/05/gig-workers-survey-san-francisco/">gig workers are usually people of color.</a> However, on Instacart white women workers might be more present because they need a car to be able to buy groceries.<a href="https://www.shrm.org/resourcesandtools/hr-topics/compensation/pages/racial-wage-gaps-persistence-poses-challenge.aspx"> White women usually are better off economically than women of color </a>and hence they can more likely better afford to have their own car to work on the platform and buy groceries. Notice that Uber/Lyft also requires that you have a car. But,<a href="https://www.forbes.com/sites/juliewalmsley/2019/05/29/uber-drivers-without-credit-can-now-lease-used-cars-for-0-down-and-185-a-week/#47d217831b40"> markets have emerged</a> where people can rent cars to work on Uber/Lyft. Instacart is still not as popular so these markets are not yet available. <br /><br /></span></li><li><span style="font-size: medium;">There were also discussions around how do we design worker centric platforms during the pandemic, especially for location based gig work and location independent gig work. Example of Location Independent include: Upwork, Amazon Mechanical Turk; while Location-Dependent includes Instacart, Uber/Lyft, Task Rabbit. <br /></span></li></ul></ul><ul style="text-align: left;"><ul><li><div style="box-sizing: border-box; font-family: "times new roman", serif; font-style: normal; font-variant-caps: normal; font-variant-ligatures: normal; font-weight: 400; letter-spacing: normal; text-align: left; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px;"><span style="font-size: medium;">We also discussed how the learning curves that gig platforms have can also act as a barrier for helping workers grow and transition to different jobs across gig markets. People who are established on certain platforms like Upwork might have a harder time moving to other platforms like Fiverr, etc where they have to rebuild their reputation. These type of dynamics affect workers' empowerment and where they can move to. </span></div></li></ul></ul><ul style="text-align: left;"><ul><li><div style="box-sizing: border-box; font-style: normal; font-variant-caps: normal; font-variant-ligatures: normal; font-weight: 400; letter-spacing: normal; text-align: left; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px;"><span style="font-size: medium;">There was also discussions on whether the process of designing for the worker can align incentives for all parties (platform, workers, customers) or whether it would entail concentrating power between two stakeholders (e.g., customers and workers) to "overthrow" the unfair practices of the platform. It is not clear which approach is more promising, or what are the alternatives? </span></div></li></ul></ul></div><div><span style="font-family: arial; font-size: medium;"><br /></span></div><div><ul style="text-align: left;"><ul><span style="clear: left; float: left; font-size: medium; margin-bottom: 1em; margin-right: 1em;">There were a lot of reference's to Michael Dunn's research where he studied the characteristics of the different gig markets. We discussed his characterization to start to think about how these properties might be best leveraged to create worker centric labor platforms during the pandemic and beyond. Below some nice images of the work Michael Dunn has done in this space to characterize different gig markets. </span></ul></ul><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;">Finally, I also co-chaired with Frank Morgan the session of:</span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span></div><ul style="text-align: left;"><ul><span style="clear: left; float: left; font-size: medium; margin-bottom: 1em; margin-right: 1em;"><img alt="" height="616" src="data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAmkAAARiCAYAAAD/ZmArAAAABHNCSVQICAgIfAhkiAAAIABJREFUeF7svQ+MW9Wd9/3NvmnGaMPGzabCaNOM+25QTEqVyytKXCnd+KH0jVHzEvdVAl7aLq7aB6YruvH2UcGofegQWMXtPguuYDemfd7FfdoSo0RPDEWK2VIwaiQMRYrR0sSIrOKkkTAiJEYZLU6I6vd77r32eCYejz2TmUxmvgeFmbn33PPnc+753d/5/X7n3kUNJiiJgAiIgAiIgAiIgAjMKQJ/Mqdao8aIgAiIgAiIgAiIgAjYBKSk6UYQAREQAREQAREQgTlIQEraHBwUNUkEREAEREAEREAEpKTpHhABERABERABERCBOUhAStocHBQ1SQREQAREQAREQASkpOkeEAEREAEREAEREIE5SEBK2hwcFDVJBERABERABERABKSk6R4QAREQAREQAREQgTlIQEraHBwUNUkEREAEREAEREAEpKTpHhABERABERABERCBOUhAStocHBQ1SQREQAREQAREQASkpOkeEAEREAEREAEREIE5SEBK2hwcFDVJBERABERABERABKSk6R4QAREQAREQAREQgTlIQEraHBwUNUkEREAEREAEREAEpKTpHhABERABERABERCBOUhAStocHBQ1SQREQAREQAREQASkpOkeEAEREAEREAEREIE5SEBK2hwcFDVJBERABERABERABKSk6R4QAREQAREQAREQgTlIQEraHBwUNUkEREAEREAEREAEpKTpHhABERABERABERCBOUhAStocHBQ1SQREQAREQAREQASkpOkeEAEREAEREAEREIE5SGDxTLdpeHh4pqtQ+SIgAiIgAiIgAiJwyQjMlK6zqME0k71atGjRTBavskVABERABERABETgkhKYKVVqxi1phlooFMIPfvCDSwpQlYuACIiACIiACIjAxSTw4IMPolAoXMwix5Q1K0raihUrbEVNSQREQAREQAREQATmC4F//ud/ntGuaOPAjOJV4SIgAiIgAiIgAiIwNQJS0qbGTVeJgAiIgAiIgAiIwIwSkJI2o3hVuAiIgAiIgAiIgAhMjcCsxKRNrWm6SgREQAREQAREQATmPoGJ3mQx3V2fUtLm/tirhSIgAiIgAiIgAnOZwLLBsa37z/eAj/5z2i2WkjZthCpABERABERABERgQRNYd+fY7h/aA5w8PG0kikmbNkIVIAIiIAIiIAIiIAIXn4CUtIvPVCWKgAhcQgJ//OMfMd04kEvYfFUtAiIgAi0CUtJ0M4iACMwrAv/wD/+APXvoalASAREQgcucgJS0y3wA1XwREAEREAEREIH5SUBK2vwcV/VKBERABERABETgMicgJe0yH0A1XwREQAREQAREYH4SkJI2P8dVvRIBERABERABEbjMCUhJu8wHUM0XAREQAREQARGYnwSkpM3PcVWvREAEREAEREAELnMCC1RJ+xCZ7T/Aoo0/QOCJd9whfAfx251j4f0f2sfKv9xl5/HufHvyYT72awSYd9GX9iHfNfdHyD34kFPug2+j7uat7n/SPmb+Wb/k5yTsxHZ+18k7eqxr4TN2svjYo+zbHvwwxZ+mnd89iNqM1dZe8B8w9GXD5SE0x4Xf2mgxXLTxUcR/7+Y/+zaiXzJ5/xFDzWM9t9GwdviH9p/p+ar+Mn6E8v5fIfy1f4T3iw/B++XHEPqngyid7a+Umck9OifCv3Hu/5mpZ6GUKhnT70g3ZUyu9Cv4jIz54s+RdedG7Tc/d+TO7ftRNAX/3s3Dv0uTVkR58T1Hjgb3nO6e+9hvYdlyeBdSzUcDmjKIxykD8835+vv98Nsyfw9y/c7h5rVf/pXTn+6tmtrZ948i+eBP4f/SQ/B8cSf839iN5MGZkm19NrGv8euz7HmWfYEqab2NoneVH1s2BBBZu7S3C3rK9TFYa312zg+OV1Gxf/sIpdeqravLh6qu8nYSpePneXwpr/lE6/zs//IOcq/VMGAF8H/9H7Nd+wqE1pqvl51H6dBJt/IqCocMF5NqKB5xBc+pKsojPLRkJUKr3dNz6EfttzmEkq/jeY6pb60ffpzEy8/mEH5kVFmfanNrvz+I+HcfQ3iyh9BUK9B1M0JAMqaJdVTGhD7WA+qlKxGhbN5y40p4e8jec5bBlbBscX+ScuUj57J3TqB4yi1hhLLnuPN77fhJ2FKbz4ngQM81zFLG95C89xe4/8UTqC4xMnQpqkfKuP/ebJvyOcWmnD2N/C/3IPi1X01ikJhi+bpsDAEpaV1uCN/nb0HuH/4amS1Xd8nV/yn/2pWwP8V6/ATK9grsJApHmjY14OyRE87q8AyVNCMFlvgurdLxTgX544sR3ODHlf13d5pXXEEFdYVdxrtUam2h+H4VxVGdFmVCMvTqR6ikmfOr56LQ/BD5Z8t4l81bduNmFH/8NRTuWQ0j2999rdyDNcB0bKL0EQpPPYcfv3ZylqybE7VDx/slIBnjEntnVMb0pHQNXo80ZXPuv32Gi52LmShrW4tCx1dQO1Rx5IpdjVkUup4WLhqN+B7kottZdl/MdkyzrLdLyBwxC9kViP/oW8j/eCviq/jnuRPIlaZpTau+jvhP3sSrp5oL5Wm2VZd3JaAPrHfBY9yd1/6kimXhr6J2/zVUDg4jvuMFZHnz+2+6Dv7XDuDp6ko8mv2viLfKGUH+iZ9j6NkKar4AUg9sQ8zWyNoSlQhrSRHHznG1xlVZZDmVDrM6863Eek6iV6tU0t4HgvxpKx1mpUbtqP727xB75ADyR0ZQX7oUgbXXIfmdLyL85ybTe8jszGH4AK1Na29AhEb0H792HpseSCD/hSvY9qNIPMK200pX5drTWhvA8D281uifZ2gW57k0rXnVcx74LQupRLNcoEol4g2KoUctNuJIe0fG/W7KSb6ADOuo2HVYSCY+j9ASuiH/31/g6RE/dv3vr2OI7S380z/ivzxLs9fqm3H4//s8AjDu5jR+3OTZphcHqNRexVa/S+W1jM/CS6FpXBzLeNx36ATeOmKE6DXwcGVrhOZVq1fagtvhVeTKt0ZeKxC86WZk7r6GPTHujyS+fOA8Nm69Gd7S6/DGvolQe3fOHsXQ32TwRHUp7vxRHJn1H0Pt969g6PHX3fJ8CN4YROqezyAw0Km8e5D5fLtJoEaF2xFq/ubqn+MU37oCdVr+nAfTGeQeM2PIfo4shn/1asTv2Yyha1iOcaf/zQG8RYV9k3UexdIIPLwvEt/ZitCBtN0Xk159PAV/9ZsornoBVz9SwVUbNiB2roz8qs0ofXsl3a15DO09QgttHd5VPoTviCD1hY/D0953/T5rBCRjHNRjZEzTatVtFIy77G9fx7u+IA4+fQss443Yk0OM93Z1iR9xawT3P3vCld3+Vkn1I68j9o0SsmbRuTWC3N2fGmeJuwLB1ZyNXPBUuCis4xNcBJ6gXFmM9dYKvMoFoVkU4pYVKFOugMcty6hoPczdVRbus2rIHV+N3F1jO1d95kn4OV+xNozyrs9Rfp2mLHiuTRb4EYttRuJ6yvKmLGgv78dGho6mOmUhS6MThkqnUc7Yy8jWIOp8zvh9zmO/F/k4aK2GzzyPRjywKD+z9wDxb1IOmSJHSrhl4wiefG4DikZWnlqJu+/yovTseQz9y18jtmSC58Hsr/RHwVyGvy14S1rlQB6RB/cg+r088pMIh2KG1orSSXg23Izkqirybdac1tiPcHIsDyDMifHBkTeReOrohbfFQNMyxgf38Q9pAeJDk7mWWUEMrTYTyFjWPqJ5+qRjeaHSEaASlko+h6cP1RG866vI3eRB+cABRB8/7FiRfltAPH8Cx3xUsLaCD/H2Vc5ppHf8Aj88QPVsQxipDR6UeG0keZCqz4fIJrM0i1fhu3UrsrcuxVtUPpvlmri4PJUGrGafuhoUzyDzfZrXWYfnxpsxbIF1vIDwDtbB/obdfpVsV8F7KB4yfkmm4xUUzcLuTBUlw5+KR2h8PbZSy3NkW3qHwtgVmsFbgwga14QRImcY72W7Qyl8KUxx5jBiccPrPIJ33Izo8hqefypLd2Az3s+uHS/vfQHP0Io5asd0jlfyBWQ4vgNWCMNU0Ozy7s3jaSqpIQrLIR/Le3Yvont7K8+4a+vnnLK9S12V6Oprkfz2LUjdTUWPp8q//AW+zIdMZTmV7ztWol4q4Vv35kZjYMzl584jsHWzzfddui+GuRgIhMPYvtYpe004gsytjuXRHHmX4/xDuqpr5FL/fR5hultfPrUCQ3dZ8J+q4IkdGST6jt9z6tL/eyMgGTNFGXOuitQOymbK59jeTsK2jf/7Jdu688YpHxc2FiqUEZ1SmaEGoZu4uDtXw8tPcdHKxfD45LdWYhkPGo9Gma49R1b5EN3qxxoetxeLZ50FNlfW9HJ8rLe5e7yEH3K+Vlw50KqXi9vhpypUBL2I3XODvcAsPeHKAt8NSMUYu3HoTdz//eeQazeCTVQer6+fq9sLVizxwGu7Yql8bqGsobyJG0WvR/lYWxJA8h4aJEZqePXZInKnuDC8J+B4gqgM3/dwCOGW0nUCT7QsbF2eB62O65deCCx4Je0slYRnXnwTTx+o4K3xk2cMwdMo0YIF2hxCN30GofB1XL11SIyVGNr2WcQ3OA/KGifLhenjXK0Z7eK8rViUuDL7wFYuVjsKBlUGc6xMJc1ROsxKzcOV32bs+k4EyVs/hfAGx/rygbtiMnFsHzDXIK1gkfX8Z6+e3HSWCqUdw7US8bs+ywf0DbbV6OyhMopnz6M2Ytpofjqry/0PR5HZ6hrwz1RoHj+PNTeuHrNSGy28WQfN6G4dQ3d9jhM5aPM5W6ICevZKBNcaW9EIV5+UMmdO0H1KpXQ5GVAQF411zgg+8r9qbYd6rmyuBk2cCBXbQ2YFS8XPokJnlD9aJAvkWLTj92gxW0shRMW3YIZr1Q1IfuVzVFz9dC2exxtUOM3VzTRghXH43+5H9vNtRuWRiis06SqgwPebzCyvaJdnYXjb9Ri+w5QHvPFaxXHBugWOltduRTMnR5W0zlarMyjYcYmLEbojjPhXQoiaMTxFVu3WS95f4fUcf8vwNML4PDyD5LDcab+XCm1okP1vJi4Ydmf/Oyrf/iQfXBUc43FjXUtuuwXDN5p7sIZCaayiOXqxfrsYBCRjpipjRvDqAcpmyudnmou6iQaEoSMlI7+5mIyuvxaxG535MT67dVMQsa/cgKBZ9Nkyb3wOSlqzKDSHT1E2c/7ZMWj0cgQpb+x4NVNX9STKZlFpjl/d49ylEnbnjxKo7xpr9Sq7C8JlZnH7aSM3zlAxdGR/+I4Qnye0hhtZQLk0RhZMUJ5penNBiCVtcs2caKYe5WOAcj/0+YCzSLZlGGU5F8G2DFvi5blPjnH1rrvjmzj9668htrTb86C9Ifp9MgILXklbc8cQGi8/yH9D2O7qJZ2hNSf0YnjMRDUrFHuij0/OpPB0PDeaN8DVmnnIH6M1JGdPSBPceQVdmD669oAKFSjj1jRmassoHYwGC3JBVXyR1qnN3KkUp6m/rWpH0WJu+2HNtpk2NhOVsJqtgFbw9c3cjbQ5h+fNn3zA1zjporEg1jH/W8/+Av6/zSN3bgXCn3ZcYHWu4ArnliK0Ybx5q618uyy3DsOlnQ+PV1l3gC4+0y+jlNYOUXFjmRHWu56KW/HQe7Zr4V2jqN7o6+B6+4Sr1NbpajBcWJAtHK8kG0f5K732phO/t9wc53nTZ9MuWq2MQPG4P1vHzTmmAIOPA/ZKczRVnuUKm/rSsptuRsK4GpnqIyNOeUdewLXc0XXFvW86K9VmPe7lncpzTvG+ce+JTmo7a3AfGIvJz9Tpgc8ewxFUOz1IJrm/3OaMsYDWqk7NTUued6lzr9ZOdW5Rqwz9Mi0CkjFTlDFLVmP3vxnZ/CBOP0Dh1yXVOQ/tu9jIH/7wTKScNMvoNn/+nKEMRiniArLwIhdnlF/LuNHHupIbkprH6bUoM8uArdD1OHe5wIqub1tAmbacqyBpW9F8XNh+xlV4muWdx9P3Psydpo/jfnuh1pTjbic6leeeGn3+tHtU3JPmR4/ysXVFN16tTFTaNnzScR9P8jxoa4l+nYTABGr2JFct0NOOBYYP24vQfy8nfYDT/I0jDPA0CpRRLkzs2nLj2ivhebpKM6YeTsSQUR5NjNS9z+Fnp5Ziyz1bMcQ80UfKtvVsTOoknDjBnDavxM5/iSDSUuAWw0dTtff6W1DKXofMT15AIm9cYGlapmKMYfokirQwfrB0NSLdZaTdBKcOCgVbIWz+dI8zvsruF5Wxwmsn8AHjsCI3+RnwDqQZV1ZcahRVWomscULM7ZyJ+xigdvbGi0Vb6XOEI0/y+LKnTvL4m2OPt4RKU0hxFWjKGidsOo2ld7kXNRNvQiW2qb60hP6qIPY/QJdEs5wlS21LmxHYJnUqzzkzqqQ5CjX7+c6/I5Ys8tqVGH74BrcE/rD5NX9O734boOLbSu690Vxlj6pmEgOjkC7tb5Ix0+TfkzIxWR1cMK/24IfHGV/M8AOzGN7IRbWHC2V7UUhLfo7Kmzm+nvLHY2RdM3Wdux3mGeWHj7LvLQbh186Z3aRmgdac84ux6TvfRMoavc5rngVcQDqpQ3nuGQ+VVbP2PGuUMeP35B/FJ55EnF4RX3gzcqYcO3WXj81cPf9s49/1edBzgcq44C1p/dwCjuXMTCZe1bJO9VNCW96ruVpbbsoZwbtG6aCbL2hO07XHec/jNbxLC4qtjJjZZl4vYczrVJiGtl2LsG+sOtBUIuq2VaS5EnProyBoWv08yz+BwJITSOzIIrKjyFi4M9xOvQ+xx9+EZ+vXUX7YosXrPN56scStB3/gqzdGuBsxgNA4a5Nb8uiPpmXRWM6M5adtpeY1MQtchYaNone8jHSJr/PgxoUg+xpe66Hb1exEYrupyIXsTRAXJi8tjH5z+BR58UeQljlDYNQ1QV78O0B4ZiVtLGj2T9cC1rQ0DjSPm3MTJO9NdDGwrR8wXjHxqvvOMG7UsOUax8p7zSfgO864ve9nEc1UWorcBMW5h2kRdQN2TbyMeRjXjx+hK5kbRcxGEFtYm6znyc8I6xFUbAvaOKto90q6nvUud+4Zh8VHrpub/Rl3L3UtRCdnlIBkzNTwthZRI87CyljWpp74miRbCDP+jPLGCaFwAq/sxaI5bgs582qkj9s/pz53GVIRCzAG7iRSj7zuLvY4582zwaSlXgQGzyP/COU1Y9Kydkzv5MmzaoUjL1uvDGFsHcMdXuWCuMKwlunIx8lrZ47Jngc9FaJMhoCUtJ7vAz5kV5mHHFdXT/0a6adepxIznWQCTkdXQmbyO49QZxXXLNnf3N5N647PrFI46fKvHkWaFi/biubGvPlXr7CFx1svFpB65gAy7ZN5wMRvmbpOIPX4b5H8idkhynf8LPfRmrcYtdKb+Fm+yDis31Fp4E5I5hzwUSk6Rtdi1Xn1xliVkBlKBYS+sQuW+fetfciOGCuYqaOKTOYVpDKvuzswGc9g2gnG4ZlVKOMqnqfp3uLrPHxcNRr35sAp59hVtnVxguTjDteWBdDEnbnRqnRNuPKUF7a9T46xKSGT//ibGN7zOyT2GpcCd2Ld6HeUtwmqsQ/zvUKJO4zgHMHPHi8wbo/HjAJpBGeV5T3xChIZBinT2uYhd1sZnDRdgfCtAdvl+8GB5xD87s/5zrQ37TFcZl1HJZiBvWRhlLQiNzOkf3kAuSr/pGXVVm67po+13DvlAwzuPeYqluOuCTCmxhhrP+DGkMQzBSRfMyNNF4VlHjRKl56AZMyUx4DyypYdR4q8tyl/uLicTvJR7pq5YicTB+rG+JpFYUtGtd7HOJ25y8iNDVwUsvyzh7j5y36R9hUIcRFqYmgLT+WR+iVlOjesvTVChc3RHSfv2jUWYu5mrdS9P0V4+y8wfMhcxlCWGz9hh0FMXT5y0W+KGjmBzJ6jY2JyWw0bmOx5MHkXlMMhICWt5zvhYwhzV9+dVDTqxvJzjgK152s7ZeT7v8wuRDt5qXQ0H5SjqzhjRWnlGVjNXTYBrKFpPEPFobLhBmw0lxrfP3/4GD+V4q6lQSpBqRepjLhCxSn/SsQSUWzfwBlO5WqYr+YIcIdqLnE9FaUrEOVrHO5mIHr1xef4eoaT8PHVErkHPstt6Ef4yge6JTu5ILk76g2+l8z+xzizKmPbYg98FTu5YaLGXZ0J1mGxjvwDzTgLKmburik7/s5VDLytjQLckGErKRMkKpotpda4gFv9azvO11O0XmI7cA0yqc24ne88Kv4kjyx3NG7aGkVuW28KifcLjuDE8dfJhIH1LC/1wM0sjztjqURlT3lZHnfDbqPA6zF5Px9BPsFxW7UYVa5qTVze+ps289i1ttALfOWr2Ld1NXddcqfaUyfg4StWdv1oM8KTWTF5bSjGtnEzSp2vExk+MMED6tMcD1P/0irSfJVIeakfdz8QQ/KaHjugbDNMQDJmyoAHuQvyrgBja0eQ44LMY16jMZ3EXebNReGAiUdrzsGmB8SUbfK4x6czdzFwNRK2Na2O5ymrzA5O/61mlzY3UY2UuSCs8NUclAUPh9t2Uk7WuU8g8aModto7Wat4nhsvPKtWY/vDX8Wwme/TkY9XX4dhbsQaXFrD01yMdzZWTP48mKwHOu8QWNRgmkkYixYtwlY+zPbs2TOT1cxK2bVjf7DfaO9d9Um6DA8jsjmLZ86570mbJK5+phtYf/89Bs7X4Vm+AtbV55He/j/wrRJjGh4w70kzcQ79pjNOGefCOGq/t0dJBC4PAg899BDWrFmD22677fJocFsrJWOmOGR8VYbZfV/njkPrmitRfuIxXM9Y1dY7LqdYrC4TgckIbNu2DXv37gX+6oGxWQ9R5zl5GNNVsSaOPJysZQvwfInv2fov+REM3hRB2jI7FAmBLkOrGT9wCZnUSy8gtKOMs6tvwO67ViBn7waiK4vv8Jla4qsgYjG8xP75p1aArhIBEeiTgGRMn8Ca2U8dwRDfi/gqN+Hc98AGegyMf4EeBfcTfFMsVZeJwCUnICWtjyEIcVflzhHGCDCmKPKa82b+XfeEJw+q76OOqWb1foE7dvglhHi+hNj3uWuTpvi779iMeCuwot+S+TqQ6z81cYxYv8UpvwiIwKQEJGMmRdQ5w9WfRSZRxRBfpprasdf2KGziS6wzW1pvWu18nY6KwBwnIHfnHB8gNU8ERKA/Apezu7O/niq3CIjApSYw0+5ObRy41COs+kVABERABERABESgAwEpaR2g6JAIiIAIiIAIiIAIXGoCUtIu9QiofhEQAREQAREQARHoQEBKWgcoOiQCIiACIiACIiACl5qAlLRLPQKqXwREQAREQAREQAQ6EJCS1gGKDomACIiACIiACIjApSYgJe1Sj4DqFwEREAEREAEREIEOBKSkdYCiQyIgAiIgAiIgAiJwqQlISbvUI6D6RUAEREAEREAERKADASlpHaBceKiKbNSCZfFfwAePx4eA+Z3/Ytnqhdnn0ZF8lB8sTlUuXo/yUXi9fptfIOCH3x/ip6zMd/aAYjyASLY+hbrqKKUiCFrBeT8eU4CjSy4nAvUSMrEgAn4zNzhPQkPIlHuZExWkgiE0p+pFn7fFIfi9o3LPyL54oZd29QJ/bNt7ueKi5+kil5CPwR8rTKFKyaUpQNMl4wjo25093RI+RLMlRE3echJWzINcMX75fHi8nEYsZ/HbdsHJezsubyhdQv4i99QTTqGcjdhtqZOjFR1GsZKavG3NHBf0p4Bkyovhcg5hT+/FKKcIzC0CdS5Uokh50yhUQvCxcdVCHOFIAoFSCsE+7u2ZmLcIJlGkwuKdW9AuWmsmkks9SE2nDfUikkMlRDJD7jePJZcu2uAs4IJkSVsIg18pUgnqcdU7Lq+HVi/fDEplTzCMYL2MimNM6200xvenXkPd44W3j4dYbxUplwjMJgFa0fJ+xIcdBc3U7AslEPflkCn1146Znrf9tebyyz0luUQ5VihV0JK0kkuX38DPwRYvbCWNk8q4Fvx8unsDYSTylb6HqJQIIJzKIhG2EEmb66vIxUN0hQbpfgsgGMvA8VaUkLBCSKQTiEXM+RCGck1XaQ3FZNi9JshySs5Epwk+EEshFQsjFLQQjCRRaCozdttDCAZZD90csWzZFQ50HYRYTzaJCI/ftyOGUCKPSi6OUHgIdpU1rvDYXtNGK8B2p5z6KpkL8451QbKdqahbZxDheA4OsW59mwRpvc52U8G6QBHszPGCNv42g2h4mJaHLIZCYdd1OlE7x7JJ8sFX4PjHUkmWQWWRCmMiX+L4RRAOkU04Mcqb5cc4BhZ5W6E48vPbyz3JoOl0zwT6kTE1LlZotfaPmQs++ANAxV7F9D7P2udtPRehu45yZCiKsLmHIymUXE2ibuZNOIgQZUYoFEG8JZN66+GF8s/IsijllSkvhHA03Xdd9VIa0WCAc40hERbnZFPoVdMIWUNIJ2OIsM1WMEZXsNvOSg7xsKnTzE9XzplTdB+njZy0+xdCJFlAT+vBPuUS4y0QoXwuliknKIcS/+unfcgljuy454jkUm/334LI1ZjhRIiNrVu3znAtUyn+w8Yr2wcbpn2tfwPrGo8enqSswzsb69Y/2jjqZjt435rGsnXbG/vfcQ6c3relMbhld+O0/ec7jSc3XdW4fd+H/P1g4741yxrrmxWwnPVr7uNRk21XY+M69/cPX2rcueb2hn3J/tsbAwMbG48eNZk+bBy8b11j8O5XzB+No4+ub6zb/gqPmlOvNLava+Y72nh0/UBj8PbdjaP2Sab9dzbW3P2S+0ej8dLdg42Nu446f5/e37h9cGPjSbf94/O+sn1NY8tut6BXttt9dxCdbuy73ZwzPe3St1at7i/s01W373P/ON04uGtTY3DTkyTV4HiM1jUxR2Yc15/Gh7sbW9Zsbzhk7IImaOeFbF6686rGVRwvu/tmbAcGOV4ODMN7vQOxVkJ6AAAgAElEQVTfZub01QzXxsa6nZPdKOM7rr9ni8COHTsaTz/99GxV16WePmWMkQPt97FbspkXG3eZe7LbPDP3dlMGjJ1LH1ImDQze2dhvCyUzb5v3spFP6xvbX3Hn9ztPNjaZfE250d6zV+6256kj10ZPjJd/H5q515xPlE6m7c4c6lZXe9sPN+5bt65x30GnEaa8Net2OjLH8BlY02rvO09Sdty5n7WwT1tMn5x5a4435d3hnesbG9053PjwcGPn+jUsu8OQdZFLRt4M3vmSfVFXuXSa/Jpy3GTuWS5xZMc9RySXOozRHD1k9Btbh/irB8b+W3GtfXy6aQFb0srI5o6NVcTPvoHsFKxp1lACYRNAwuQNZ1DKMjje/stHa5oPtZq7dvMEEI1wWWxSgCs+rpttgwxddZ5axXH5eYJI5oYRcl13Xlq/hvx2JsZuheApl3gNV6tcXYaiQR41p4JceVZQKDYN7awnEYXfLcNkaU9BmpByTqFssIWQv4ZqD8vLSrEITzTixlt4uUq2UKLlyU4T9W1c3ebPej7ubrzgirMYQoYWPBdfK3dXjh3KbD/UtZ1s/Xg2QfbJrp9WxYCXLN3BtLiar5Udk5nX40G1UrWtlb5ohhtJ/JO0QqdFYCoyxuPM6TZ4Y6ZxH/Osnb83SCuaLZS8tCr57XuZAaHIV2hJbwa7+Whp8pfh3vIXDl8xQXnmbqCyYsi5MqNd/lUKBfho0XLms4fW/yCqRZq7eq7LT2t2AcOW02sPPQ4BysaWePKHKeuccz5a2rxVMycrtGD5af12avXRopWKBdj+KuVkHeGw3+kL2YUpWEulzsLu0soloJ2jabDk0oW34EI8soA3DnCnkn8AOHa2bdyXcUfVeHVh8tvC2+6ro8JFLQ2pbAElCsIKYxSQaJbR7tZzBI156MMbQWo4hwjdj/lkBqlooCWoPdxR1RLSjA/z2mZ4o1R50d5UL30ko8qgt2scmYfu3Uo+TWWwiAqFXIkyNDZ5N1GjJucN2JLeTh4TrFYzQtIcm6Bvrdyjv7QH6HY47RbejeOEV9knJm6n31aIx8bY0dXdHsxGZaz1pw3eHiFYCbpaonRJc6NDOk3XMYtSEoHuBKYiY8z8HpvG/t37PGsvxdt+0/MedxIlSTWPBN1zzj1PdadcQ2R8A5oFddg4UOK5dvlXr3FHY4YKYc6Vb9Uyav4I+9RrXWY+VlFIp7k5q0T5VEYRwdGuUB6OSiAubu0zNdTqbfOWC096Q5mo3FHByzBcpGBfxDaUK/AkOnfw0sklp3tjniPsmeSSw2Wh/39BK2kmFikb+nu8/IFzGwxuoeISaalEU7o3ajnGbCW50zAVxzC1qFqG1qJJS/Iw9iyLUiiH4SEqAvkUChnXujPptf1mqKMQDyJeHaJiOAy+UQS5aLS3OI1+q5pG/qlxnEaFk11KK0OywJ1bVNCGGJcWyxUQd1f7k12q8wuVgM+Od+xZxlB58tiLsLGpzmNmYTUTyeOLIl3gjvWLWHhwOE+ZMq699SwXdb3UVUIyFEUhmkJymK++8BY53/I9tI71dULksTCUySPu76GIHrJILvUASVkuKoEF7O7knLbiDDg/jJf27ca+V46iTAVrunO5SAualUjZga9+29JUm3zAKgVki1V4uOJM5rIIF1NOgD+vrNmWKjcZsz+tduY/n6fGVeZo0bUKrVwXRt93qJuuzrwP8fQQQnxPmWmieQj0ksxqvGWt4wV14yNtt/T1UkiPeabE0S374rezhlIuzw0gdBUNZZAf9iKVLvbYE2VbyAT6kjG0lPtp/eFUbku0xldAC/+o/eji8TRKYdUJubALraGQSk1rU4xRJqu0zrcSA+kT9tbUHusqU8GjXT8V52KVi1yvh0rrpOKJZVMetkRttYhcgdBMnexdeyhHJZdEuhUWMtrMXn+TXOqVlPJdLAILWkmzIXoDCEWijMvwd1yI9Qva66Mrsak91fJIZyuTC5kqLWhcfdqizaym214nUS8w/sk+QTcCFUAPXaJGSbMYYFLIubtAGe+RLfq4o6rTUtLpgXFDOLLOuPsqdHE6f9WLaWQotHi6lUbzju29nxakWq65o7OGPN8dZ4UDYzNdpL8m4zhRG031M9HOEi1o3HRrJ4/tEp2Jh+ZFgqdi5haBnmWMiYfkDuRksWXZrtHKlaqEEbNmoEu0MoX8BWSawWWUQ0nu9G7zJ/ZdqZ9xYtVsFmX7Sr73LW12pFMu9VqXcWfSxekoqnWUM3xnHGVTdz2Nu+gDFeQLjnJYzQ8jTrlrxwQH6SnIFJ3rudM2y3CS8jQkfVe5NIEltAlRcqnv20kXkMACdnfOzPgHaUWzGIgezNFK5eMb9BncX5rMmhZMIOmnohimS4B5vUNp0FhjJ18ogNIQt48bJcsXYzyUoxQFaNGJD9EaFqL8owTzxzPuBoMO/QpGEIrH7O3syTzzJSOIDHHbvY8xboEIuB8B1aYUHJfX315ccBiZyBBfReFU6rG4Ko36mKNt5dyh+qkcmpgj6xvXxuh4fWnCdlam0hRe40U0yW39HIew1yi03NyRnYmn5hSbp8vmCQEPQnydzxDnteWnZd30im72RDbd14tse4fhRyzNF0nzFRUWXwbtYVxXMJ2Z1guhPaEUMnRXRq2cYwXzm1dmODKrp7roEk3y1T4xCrYUrYf+MF+H43GVrAk7xvlJt3KBrxgJpTk/GcOWzobs3NYw5V00TmWNCyuqat5IBhkqblNNXeUSN3BFAsOIclNYJFXgq5DG1TKhXJpqaySXpkrucrpukdkeOpMNXrRoEbhFFXv27JnJauZn2XxPmj83hEp6/Gyfn91Vr0TgYhB46KGHsGbNGtx2220XoziVIQIiIAITEti2bRv27t0LvoJjbJ5D1HlOHjbv4Jjw2l5OyN3ZCyXlEQEREAEREAEREIFZJiAlbZaBqzoREAEREAEREAER6IWAYtJ6oXSp8jBGjTHDSiIgAiIgAiIgAguQgCxpC3DQ1WUREAEREAEREIG5T0BK2twfI7VQBERABERABERgARKQkrYAB11dFgEREAEREAERmPsEpKT1PEZFflrEA5/5CDff1O/38wWK4Tiy7kthuxfDF1Ty3WKpipMrzxd7Wc0/ul94kc7W+H6zMNLVAt+lFkF2zJsha8jyA8RDhYtUVa/FtD7WzC8zeBfBS56W+XhzKMlv9fWXinG+jy7X/XWX/D4X37fE9y5x/KwA62R9Yb401L7KnLMScN9V21fltUKC11oIxvOTvHCzr2KVWQREQAREQAQgJa2fm4AvKxzm9xvL/EhvpVJEOlJBIjLc98M9xFfX52P+fmqeXl6+STxX5wtszXtn50oyH2sulVDi191T4asQTpf5O/+m0jONd0127x1flJkts44y32heSsHHT2Olyt0vGXuWX1mIUzFvvbu3hlwyhwA/e1NMhafxHvN+2qC8IiACIiACC4WAlLQpjzStYUMpxL15Pvj7K8TDb/SZb2bOVqrms6jzKwhzSUebrb5PWI83iHCgCurbfaQaysVS28fozXcFzXdU+yhCWUVABERABESgRwILW0njt9wysSDdbfxkSCCMRL7SI7ZmNuP2dD6AbI7UK/ykS5ifWwrxM06hCOK5zp9Lst1zxudYzyPKz0G1e+qKQxZieXOuhHQshKBdVgiRZMFRDupZfnpkCJl0FEFrCD9LBWEl27TESpr5U/xMczNV+Y1N8JuAPapoNhPWy4/eBemijVEDZWuASgqhcALpIX6mJcRz0TSK/O5nLMy8/J5oNDNaY68cWk0c84v5Xl+MZdKFyO8ABoL8HFMTo2HCz2xZpm3mE1fF2gVFVLMRBCLZnj5UxY/E8KP044uYoP56AclwlJa3EpIRjkfqf9ECGEGSlkAzTsZ1alKVlsCIzY7/Iua7hW75eX6Wi8wy/LSUFWP72M5gnEp+NMxvrjqfkSllhxAmT8uKIFWyqTv3wSR9dmvQDxEQAREQgXlGYAErafz4byKMr//sVRz74Cw+eOt5/DDCh2OfVjHqd7SmmAdqFVla1jzDBRQK/MeHcDmegNG3JkzmW28WXZ+tIKwyciUqcCEPyqkhZK0MiqasPL/dx89DJd2gqXolhyxYdimNO6koIJdvKWW1Qg71cBj+ZqV0dWY9UfSqo1XSMaS8VC6KRdadhJffrExXnMLqBdYzlOWHjPNI1If5PdAaFdEC88VRHx52+zoFDu2AqubboqACRhci/+WiJcSHHUD13DCSHn5r0LQtYyGbGKeMVakk89phKrCTqaS1Ir87WrZoTRs3OhPV7+E3FKntxhnTlmCfc/G/QTyfQ8IKYChTQD4RtJXuxFCJ8XFsH9uYjTAGsC1Wrcbrq7EcSpmo/Q3rUq5EN2+e90sSPrpLowW6Y/PkmfIhPZyzlfJJ+zyu+fpTBERABERg/hBYwEpaGdncsbEjefYNPiQrfY9unR8/R72IfMWiFYVam0n8MHLEX0a52q04L0IRP4p5V/uq5FH00rLiqVLxqdOq4ncu9jDInYpbqeSYZTy+CJUBy37QgxbAMJWDgl1PnQpdlda3Uc2jmsvBQ+WzpbTQIhQ3Afqtf0HEi01NssZ6awjRcmP3gkpkNFihwuac97CuqGXOeGnh8tNiFYHzJ61ePloUba1iKhxMZW7iR+RztNCxu3YKsJ2ouRCN2atacaxkDPTPpSMOA5OxbpTDJJXkNOxvvndKVOKiZuMAy4wM12glTPOD6eMydqu/U5ntx8ocP1pGm7F/flre/KUCym4eTyiOeHC0Qm8o5tTvIb+ADyFa5MxZDz/Q7K+4/ezW58nao/MiIAIiIAKXNYEF/MUBHwL+AeDY2bYBXEb35URP+M7jbBvRPEajqFOXoCUlFIaxrpm/K+UaIk39p/Pl1OUYcJ4u0BJGpatYAP1ofFCzrFqFbscwCvYznX8zeMqTcAvjg3v0UU+FI1RBkspVjFanfNkoVs3KqsjlPIhm2zQRWoRSpRyirhJEDYi7Oy2wZqYaqjXugGxD4PV72ZaaU6DH16rXdNnL3ZIXpqlxGC2HrmcqeplklhsLqKhUGAPmD9unPeFhpHIRhKwiEukUhoLNhtZRpYs06R1GodvuCHvjAK2SFza67cjE9Xe9zJys1lD3jTIyvlQvFUyXHjz8u4Wd2b3OjeIWy7zNk22ZJu7zpK1RBhEQAREQgcucwIJW0mKpJLKhv8fLHzijOLglzXij9sfoZKNrAsdHH7YeKgFpuq6syS5rP+9nzFk9wdilIfjyNYQTfp6tUCOx6EbL87Uf4wrroPRZjJEqp4uoBQoo0dqVal5Ct2iOrs52Ha2fpk0175Q4tNrM12GEM4yzSyHBHbDecgLBdPOkHxHujA0WqKDFLOQSdDM2d8lSAaq7ClEn1bHnvlS61d9zKRcxY5c+X8RaVJQIiIAIiMDcI7CA3Z1GD4qjUDmMl/btxr5XjqKco3uqrzGqcHegn64qc5EHHrrcqq3rayikUqD3cZJkLGEM7mcsUq5idhy6ZbEkGmZaqZJjDFXLLTm2SE+QQehlxp5li/BRYWuqmRXGTHkZmN670mJ2KtZQaWtzjT5M74XR9V36NFUOTpG1QhblcJLKskWLnrEYUhF2a6sWsxwv40mOMwwvjloq7b7+hO+vi6SR8qUQ6zeocFxPutXfpdPOKbbXUx21nNEEiZq3zbI2aQEXZpi4zxfm1REREAEREIH5RWBBK2n2UHoDjAtjHBFjrPqxodnuzOwwMt4YY89YDi1fIX8BmZyrWTFgP5kqmfCtSZNFd2Mpxfet0QrmWOF83B3I/QAZ92Wr3HGZTWYY2zRBC+nCjATyGM7QWxpqVlihq9PLXZ09NKDVQr5WhPkLDGi3FSO6HbNFxko14+wm7cn0OJjiPebdJHyPmaMnMs4slUeVjTHtqRdpXctVRlvR5n6lLxFhKsV+bnxobrDopbnj83Sr38lrXNHjr3L/5vgFab1sKubm1ScVKwRb755i6trnKZapy0RABERABC4PAgvY3TmFAeIrIIZDFtK2rmS+PhBBmq9N8NtF+RFLc+chg7+tFC0qdQ/ddOYt95PX4+HrOgLlGGPOqJm5yaLGNRSNU1ljjBRVFG8kg4w53TQrjSnWQ0XT4itA6Dr1uSeMsuDt39UZGMogPjTE13iwh6zLH2c7/CyzMqbCLn9MnYMp1MRgDTO+LBzM0pJGZZW7VwNUfE23/bEkNzVEEcob16YJ32MeHm8Z/rjpIsVjwRi/AsDdm/amhj5Tt/o9rC3EYL4IzZ15cirYrum25OFrQVLc0RkJ2vdI3RtCOj29l9xO1Oc+u6XsIiACIiAClyGBRQ2mmWz3okWLsHXrVuzZs2cmq1nwZdf4yo9wiW/xTzp2mwpf6ZDw0wXaV4zdgscoAPOAwEMPPYQ1a9bgtttumwe9URdEQATmMoFt27Zh7969wF89MLaZh6jznDyM6apYcnfO5dHvuW189QbdgMFI07HGOC7/EBK9mPF6rkMZRUAEREAEREAEZpOA3J2zSXum6uK7z3J8MSs3PLrJY7/DTEkEREAEREAERODyJSAl7fIdu9GWMxYqUwrPh56oDyIgAiIgAiIgAi4BuTt1K4iACIiACIiACIjAHCQgJW0ODoqaJAIiIAIiIAIiIAJS0nQPiIAIiIAIiIAIiMAcJCAlbQ4OipokAiIgAiIgAiIgAlLSdA+IgAiIgAiIgAiIwBwkICVtDg6KmiQCIiACIiACIiACUtJ0D4iACIiACIiACIjAHCQgJW0ODoqaJAIiIAIiIAIiIAJS0nQPiIAIiIAIiIAIiMAcJCAlbQ4OipokAiIgAiIgAiIgAlLSdA+IgAiIgAiIgAiIwBwkICVtDg6KmiQCIiACIiACIiACUtJ0D4iACIiACIiACIjAHCQgJW0ODoqaJAIiIAIiIAIiIAJS0nQPiIAIiIAIiIAIiMAcJCAlbQ4OipokAiIgAiIgAiIgAlLSdA+IgAiIgAiIgAiIwBwkICVtDg6KmiQCIiACIiACIiACi+cDgkqlgrvvvns+dEV9EAERmCaB//iP/8Cf/umf4sUXX8TRo0enWZouFwERmE8E/u7v/g5f+tKXLpsuzQslbeXKlfjXf/3Xywa6GioCIjBzBFKpFP7yL/8SW7ZswR//+MeZq0gli4AIXHYEvF7vZdXmeaGkLV68GH/xF39xWYFXY0VABGaGwJ/92Z9h+fLluPrqq2emApUqAiIgArNEQDFpswRa1YiACIiACIiACIhAPwSkpPVDS3lFQAREQAREQAREYJYISEmbJdCqRgREQAREQAREQAT6ISAlrR9ayisCIiACIiACIiACs0RAStosgVY1IiACIiACIiACItAPASlp/dBSXhEQAREQAREQARGYJQJS0mYJtKoRAREQAREQAREQgX4ISEnrh5byioAIiIAIiIAIiMAsEZCSNkugVY0IiIAIiIAIiIAI9ENASlo/tJRXBERABERABERABGaJgJS0WQKtakRABERABERABESgHwJS0vqhpbwiIAIiIAIiIAIiMEsEpKTNEmhVIwIiIAIiIAIiIAL9EJCS1g8t5RUBERABERABERCBWSIgJW2WQKsaERABERABERABEeiHgJS0fmgprwiIgAiIgAiIgAjMEgEpabMEWtWIgAiIgAiIgAiIQD8EpKT1Q0t5RUAEREAEREAERGCWCEhJmyXQqkYEREAEREAEREAE+iEgJa0fWsorAiIgAiIgAiIgArNEQEraLIFWNSIgAiIgAiIgAiLQD4GFp6SdOYjwxh9g0cZHEf99P6imnre6/0nWxzpv349is5hjv4Vlt+MH8Gw/iJp7vLJnV+tYdepVTvvK+qt74DWMfjd/ebVYP3h42rwmKqD2+98htv0x+L74EDxfehTWd/cj985HE2Wf1ePlXzr3mnfn27Na74KtzJ5TnPNf2oP8rED4CLkHH7LliffBt1F362zJIx63fvmee/RDZL7r5B09NiuNvKCS4mOP2oxypV/BZ3h98efInnWy1X7z87Gy9PduHsrW0gUljT9AHt9z+hjcc3r8ybF/t+TzLqTeaZ76A4a+7MhsewzdNuH3++F3xzXXPNa99NGzzWu//KvRZ0Ov1/aa7/2jSD74U/i/RBn0xZ3wf2M3kgfP9Hr1zObra/xmtilztfSFp6RdgpHwrV6JNabeahXl950G1A5VUHbbcvZ4xRUwH6F8yFHX/Gt98F2CtjarLB6o4IPlfkT+z9lvxOXIqyOl9w8ieu9z+FnpJOq+lQj56njjtSK+fG+hhwdKxxJHD1LwpnY+Cf/3/r2l4E9yhU4vOAIfg0U5YtIHx6uo2L99hNJro8u/8qGqq7ydROn4eZ5fyms+Yee8NOkd5F6rYcAKIPSxHlqwdCUiGwLYcuNKeHvI3nOWwZWwlprcJ1E84i6q3jmB4im3hJEqCsed32vHT8ImusqP4EDPNcxSxveQvPcXuP/FE6guWYHQ2qWoHinj/nuzbcrnFJty9jTyv9yD4Nd+NUuLjim28zK/TErabAwgJ29rwrsTu1SqorXoOlVF8ZhpSI2C0qx3PbCsFbPRsgnq+AMF5QiWUVAGl0yQZSYPX3a8OsOovPg6nh/hueUWsv/z68g/HMJ6k/X4EeTt8Z56qh4o4O/zFVTPTb0MXTn/CfjXrsSg6ebxEyjbAuckCkeaNjXg7JETzoLhDJU0o2ks8SG0+hJyeaeC/PHFCG7w96Z0DV6P9D/8NXL/7TPwX9Rmk8PaxSzxPEruwrl9YW1kdfHIh3aN5UMnbVk+eIkX1h27/3YJmSNG+V6B+I++hfyPtyK+in+eO0FL5TStadXXEf/Jm3j1lClfaaYISEkbQ5aWrP2/Qugbj8Jrm4WfxNBvTtsrzaZryPdPR50r3v8dQra78iFEfuustIwp3mNcC3QfjYpBZh9oCr4RKmHGzP4eJ7h5evuw0RYE7mrtrFnNmuJNfi4jz9BM/b1d8Bkz9Zf+Ef5v7UHq945gsFfEe/bAun0nfF+j+fqffuq4NVquq9PIPfZz+7y5NrB9D03czWvPIP8Ez32Z59hPH83fo+Wy+rfLyFcXI0RB6XF6O8H/z4yrYx/SbxsWTdfJKJv6b3fbbBZtfBJp25r4EfI7d9r8wr9ptsutZlq8forAl8nry48h+L3fouDKoeozjsvZ971fI/Hdx2A95o5jq2cf0jVk2vMDBJ5w/Ru0ViW+RzdBq7xfI++emrw8oHLkpFP6aneFfbUfQ1uD2L71Oldpn/h+A466rpWdCH6XFjPeA97bf4roM++hfnAfrEcqdtlnX9uLj3/jFVSO/RoBw/dr+5D4pyc53r+1LbX1t3+H6LfobrWv34XwE287q/5Wv/XL3CAwQ3OJ955lFlrnKGOMbHmfC0Lzk5bd9cv5s0olzcxHo8QZEMYadGXzvjFy0Mwluuk5b/KuF8DIr8xO4z6jjPzurxF33aStedxl3tgyreV6o1zi9aPlsjmvlfEG5V/EYiN6SRe4yyaTi06h9SOvI/aNf6T8exShJ452sEZfgeBqxzZXoRXSfgaUTlAZW4z1lo9/neffRqs9g/Jx4/1YzIW1OT7ROPLUBHO0vZtGrhg56fkW57R9oosc76G8Oq18djlL+UwxyhlV34gtg4KI+Myzp5uMGHUP+7f/HEH7WfIogv/076ic/XdEvnkAb5kCRkq4ZePPkeHYOu7gn2LIWNhu342MuWfs51hnuWw3QKkrASlpbXjqv88jnHwdL59agaG7LPhPVfDEjgwSjF0LWCtxFfO+65q263RXOjEQXGnZk7W5ouIqkKb3scrNla0Jb1Zd9TMnHFM5BWL8JmfCl4zboUolzVhGVq1E8M+B4k/24v4DVXhu3Iz8d1bDc+hN/P2OF5zYhfdL9irmjVM+xO+xUKEAaU+lJ36BL+89gorvBqRiXBrz2vu//xxyVFoqz2QReYrnVm9A7jsBeGj+/vvvu+WykErpCN5aQjeCdcWYMsf/Uf6lW8fy65C8YyXqpRK+dW+OsRoUcLYlkILMXYWWXjMCziQ+JGz3Aa2G9oreh/Da8fVMhdcZZL5Ps77N62YMW0DpwAsI7zg4Ril598AB/JDulBqFanuqv/06hl9ke+y+XM1Tp5He8Qv8kOV5N4SR2uBheQcQSfZWnim7fs5ZYQ4s9bhWgasR+/YtSH378whzfLvdb6NtYxlWCJlbfbw/TuDpp15HaXUQqbBraV19A3ZzDM1dZKfjJfzw2Qoq5j46cxix+HN4+tB5BO+4GdHlNTz/VBbhPc04pNFa9NulJTBjc6m14DFW+g9RP3LEllvLrCCGVjsLxALnY5ULinfNcYZmBKiEpZLmvqkjeNdXkbvJgzLv/ejjh21lpf7bAuL5Ezjms5DaSjlVarekdJs3HyKbzNL1VoXv1q3I3roUb702Wq5Z3OUPUI6tDiBspmAznasitWMPog/uQWzvJJG6k8jFZpFlunZDN62E71wNLz/1ArItBXS0Wj9l/jL+aayNZbr2ioechXV0q98OX3nXPu4qv+7CeuJxbOtP+xxtO2yUmeGnKpSTXsTuucG2DHaT461LJyqPGern6o7cXUIZZLtiKZu3GBl0C+LXU+72KCNqSwJI3nMd/CM1vPpsEblTfiTuCThW2iV+3EcvQbilV5/AEy0LW29yuR2Dfh9LQEpaG49KqQLjhbpqwwYkt92C4RtNUEINhRIfaqtXO64/rqrKPFqiUvQB4zeuYpZjR6pUBM5Q6TArKq4CL1A6HCXPzJEPmLdiFDw+RJetXU0FxXFHHKOSVuGkrzDPMq5+A6akDSHs+s5mpO+6nlatAAJmRVytoWysQ1z52godBVp0/bWI3dgekXGGAsVYcRYjfEcIQ9s2IGZWUSN0JRzhxK06E7c+Uof3xhCyD0ex7ztUSpnFrATzVExMuSEqEhOnMyjYsS20uN0RRvwrIURNHadO2DtnqZQAACAASURBVHUE2C+j1Fao1NYp9AtGwC0nL/5VZIwWmsKNimqoXSC7FRqluC9eZ2m+pzICrKSC/TkKkCCop+EsFc6Sox06JS8PYHf2v6Py7U+Odu3cCLKZA1zBc5VMZSZihM3ZKvJuefG7Pssyb0DIHD5URnGy8tyS664r0jNOIWxW3PV+a7VuKZX+T3H8+UAxx4zid+XVdmyJSQPL/Qh/+uNtiwIv7vxRAvVdn0eAD+SCea6sugHJr3yOD1Q/mZ7HG3wQmjtVaa4QmMm59HEuEM29YhZMXARyQfkB78egRXlmL6Tq9rGybfU1x81d5kFw62bKngiSt34KYd57Rrp84FplTBzbB/x70LIQWc9/toXGTV3nzXnUKHNMW2ojrGtrBPspezJb3SXGmQpdcOex5sbVtvwbTSN49cCbePrFN/GMrSh1SV3l4uh11k1BxL5ygxvOYdpzYZkeY4U0h0+RG+WaHYNGC2TQcsNXTF1cWJdNnJo5fnX3cRytoW2OtlVbzheQoUhdZhaZnzYBed3l+GTlmfNNGYQlYxelrWt7lBEBjkno8wHHKsvxq5/jQpr3j22MWOLluU+OLhR5aN0d38TpX38NsaU9yuXRzui3cQSkpLUBqVF5MclLy4fz07mxa6d4fGAlwjRIgfFj5XforizxMbc2SEWOeY9Q6WJMh+1GMO6CDkqHZ6074Tmxc1Ru3rUFIi1uXLna7gjGKWXc2AZjNjct8K9dgRqtU0Pf5I6k/zuLZ+yHvpkgnHwUdnZrzQqJPzxjJmHdFTrn8fS9D9OF9zjup+JkBHKN1wa23ow7Vy2mwnEAn4tmMHyIpvob3Ul25giVHU6yDasnifFo1rGYvIxA8cBn6w0jqBqBZ1x87NdZ40Z53xFwa26iIkd5XDFKrTnOtlxFRdUWhONS37y4YjR9s3mYdpiftpunPjZua/wqnVnOHili+ABp0jKQvPXjTkvI1y6PavPXN9ONuDmH582fVJKc4062C1b97mHzw+PG89U5Zp1S1/ut0wW9HGMgdXS9a5k0fTDX8H4295PH/Ynm8V7KU55ZIDCzc6m54DlGi3nOXryZAPIruJDyOQspLjzydviFl5sGzL1jLNm0kL1IS/Rmyp7467aVrZkcRYu5lxv56M635smu8+ZKRGNBrOP8fOvZX8D/t3nkzq1oLTLqtPYXzi3lgmScAF2yGrv/7UE0Xn4Qpx8wQnji1F0udrjOnaMdzgB/7kPQjt/iJoEXj6BIebDMyPEruQmoeZwWxTKzDNgK3STj2KykfY42j52rIGlb0XxcYH7GVXi6y/FWmzuV555syiDz3OiY+pUR3Xi1KqDStuGTjvegV7ncsXE6aAhMoF4vUDiuotNcfTgqWxOTs3IA3XdFWryqVMjWbF3NVWQFcZrv82Y1yhXVVVRuOikd9oSngvIqXVapF41ANIGpFIhU/owgeIZlZu3jRlAaU45xDRj3HVeW4c3YH65zl84LeNm1zrRGqOOkWexaVhZj03e+iZQ1OsxetgED1yLz8zhie15A4qkSXWC/QKAURmnX5+ArlSkovYjf2McOr2ab7J9u3UaQUZ4+wxV6qTRCAbcU0Q3XIXzqBfz4tQqDl0eo/tDSx1V6x9Q3L/O4MInCyG5H82fzuFPLQIdbfoAWPh9Xy8ea1xgTHrk65a3Ezn+JIOIYrnhiMXwcnua90ak8p6am4nweZ11B6MU7SH3vOWRPLUbknq8j1PV+m0CoNguf8GfblG7dG82yqOCb6zreMxMWqBOzSWAG5pKXikWAqsQbRxhEbspfTpljdhPQCmst4fw/8iYypo982IeMfDjL2CKzK/nUUmy5ZyuGmCf6SNm2no1Jnawzk8wb7/W3oJS9DpmfUPbkTThJmlb2GEq0bNs7ypdSpnbXw8a3ovPfF+UepzK72oMfHh9BnqEjRlHdaBbWVGKttVwa81mQo/Jmjps4NU+7ItRpHFst7fDYXUIZtPQk3mIQfu2cCQcxC99J5Ditbk7qUJ57xsPFqhFntgwyHgD+UXziScRpsfTxuZIz422niywj2vj3IpebrdDPCwnIktbGxLvcuZ2cleJHrmmeapPPOR7gRFxGS1Hp2RKVDi8VjKtpfDECkFa0Z8sUg0482oWYzRGzGcBMphG8a8zjTUEJuiNs19UIjpnjRlDa7gMTQ2ImDuvZ+lm6Rb2u0mDONRUA/jLiPHjNCnI0cXVrgoJNWupFYPA88o8wDo0xaVkql5Xf7EdsJ2PQVt2M4v/cik2s/uyh15E79hFXz1zNLaegvKatuI6/GqFiTpyn5cwIFSpdtsugubI2/aIg4yo0S2H8wVK65aiUmni9gZETyOTN7lZaJ9caYdQp9cmr3XJm2tG2QvS2YiU61cNjy2/AcJhtrdJqmfmDq8gsdSxxpkfLP0FX8wkkdpDhjmLPr8/wr17hVGgsrUZA0tqaZ2zeqxTuRr5Odr9N0NreDzdj4UZGbIta0wIyGiPXe1HKOZMEZnguXc0FopEHvA/epfIwQOt10HTnSh9s7ybjjN7lnLGtQeaJbrwFtixazVCJaxF25V+TQNNqXzcehpb1yD1LZcO2YDNdOG8YSvHLfYg9/iY8W7+O8sMWLXnn8daLlKdwd5TfyDAL04YpppZHoaNc7LdQvsLEBsT4s1NGqKygTHOEiTlumvmu7TYwry0xFvjJxrFb/dx9GQvw+XISqUdedzZxGFnaRY53K615zrNqheMRab0yhLF1DOt59RBDa+hyNlZ244kx98aMyIjpyOVeOrgA8kysgs/7zteQSe5CwRUo3ps2o0CFa5CPz2MMZk0840PxNSOEaLq1HBeYxwi3JW/ieVqCsNRy3J8DjKniPDbWISzhe8UmVDrcCW8CY5kGmu5P/m7eZTTw7BEnwNO4S20htRR+407gKq5UOsr4L6MY2pc6yeejcsiNA3TVJZ6hosZXZoymKxAyyhAVgsJTeaToxk0xDuwYFcOAkTmMp8vSTI/jdFHeSiXLlLuUk3m52ZZdZ0wEYw/aSnN+vZBX+kYWRgtice8LSDNWLmc0DxMjtdq5wmIblu01vNhfxlkY4eus6l/A8yXmYQD8xHFv/fKiwkeL4dN0JWcyr7BwBtjz/8vWdurLuM7RIhC6YwM25p/Dy8/mkb71vyI+yA0Ndnm0fD7+W9RpTcjTKui5ccO4eJkLQLUO+G+6AZsyJ/D8KVoi/naETPm7bclwrAUBis+J7zcjMidO5mFkbpOztI6kX12JuPMsGXuBiStc+jqePv4mhvd44TMKOBcS62/024K5tRCfuBqdmQkCI0cw9I1dzsOR4xG6K4ahGZ1LzoLnidcca0kznMJ2e9qWIiPnTHiF+27G5bxXjFzkgz3/6lFUzCLLZKDrys7HxccAH+lvvVhAigpcxoR5NNNAt3mzmPG9b+JnbEeRdde5kcWUOEBZ5j/m7CgP3uUfsxhtK7m3X7vKxd6KaM/lI5NBOLHKZgEdduPvTLyabZ00mbnJynltiVmEdpGJk0w4E4M8tKqMHx7ixoz91yF/y5Xd5XjznW3dunWNhdjqIsNdqPzd+1OUVjEmmOEsZuzDxluyqruM6FY0jEJuMphF956jsLZ1yE1P0ZTlcofiFuKhBW1JMy94fIPxUeafbbX69M3IJ27AxqVVpB/naobWn7sfiCHZtCo1YxF4pwyYwFtbmTJuS8fShlU81iXY3ghBE0xvkkWzuX2DM3l53O/+PvquHcZv3BXCJm6TLj71HFJLruNkM5ncOIVB7tq8K8D4jhHk9lYY29YszSnIf2uEOwIZgDtSRiJTYfzcddj1cNjegeO/lZsRwgxEP/46vv7Im6hyou58eDOixykoT3kmfPXGeF6Br3wV++jy9VMJiT91Ah5Tx482I2xz4Uqamy3seDsmq/neo0Eqoe7qcNB2w0yc+uUVe+Cr2LmBcXzc1Zngg8DawPF8oBnfMXE95oznagvDN9E0yPcHDT9+kGL5SsQSUWzfYIRuAcMsL8DyconrxwTIdi31z69HljzupEWxfpwuXr7/aY11A55MhR1rwWT3W5fCvRTowzeuwDI+8Ic5/h1VuoFrkEltxu18zUvxJ3m6WVdg09YoctvcuLsu5evUTBKo25uNmrKnzPXVzM6lK9reu+ilNag5/qOWImMBb72bcWA1d/IFsIbut8zjBVQ2UCYaHCa+iD98N92MFHdGDnIjUupFxrO6iotDrNu8uQLR72zF3ZYX1Refw9Dek/DdGETugc8Cr/W2o3zSUZlELk56/fgMZtHshjrYC2tXtqFpnTT5WwvrycdxfPFj/h64GgnbmlbH85yvZid+NznetazWyU8g8aModto7Wat4nhsvPHxObX/4qxg2z7XpyIirr8MwNyMNLq3h6czrVLw7Jd4P05DLnUpcaMcWNZhmstOLFi3C1q1bsYfv9FK6iAS4JbzEuK46d9ZY11yJ8hOP4fqnTmJZ+Kuo3T+pr7JjQ0wZ1+71Yvf//hqik7kIO5aggyJw6Qk89NBDWLNmDW677bZL35h52IL6++/RgF6nO3MFrKvPI739f+BbJca/PpBA/gsThS90A3HGKeNcGEcZF+vvlnWyczMgFyerUucXNoFt27Zh7969wF89MBbEIeo8Jw9juirWAnZ3XuY31im6TPgOrFcZ13XfAxto7TFrXK5qaZWbajIu35c2cKUtBW2qCHWdCMx7AvXSCwjtKOOseUffXSuQO2K6zLAQ8wLuKSW6fGMxvLR81KMwpWLMRTMgF6fcFl0oAheBgJS0iwDxkhRx9WeRSVQxxJcGpnbstVe1m/iOr8yWqWtYvms+1bsr75J0WpWKgAhcagLeL3BXID81FM+XEPs+dzvT3Xf3HZsZxznVlvF1INd/qmvoQ88lz4Bc7LluZRSBGSAgJW0GoM5WkYEv/D8o8J+SCIiACMwegSsRvvtrKN89ezX2U5PkYj+0lHeuE1jQGwfm+uCofSIgAiIgAiIgAguXgJS0hTv26rkIiIAIiIAIiMAcJiAlbQ4PjpomAiIgAiIgAiKwcAlISVu4Y6+ei4AIiIAIiIAIzGECUtLm8OCoaSIgAiIgAiIgAguXgJS0hTv26rkIiIAIiIAIiMAcJiAlbQ4PjpomAiIgAiIgAiKwcAlISVu4Y6+ei4AIiIAIiIAIzGECUtJ6HJx6OYuhUAB+fwABK4hoqtj5o9Y9ltdXtlICViQ7Y/UV4wFEcvXuTaplEPZ44Q9YsAIOh3CyyE8BM5lzVgKl7iV0PlvNkSvLDCUn+EBv58t0VAQWAoF6NgzPIi/np/ns22iqpIKwkuWFgEB9FIEFTUBKWk/DX0IiMgwkCihXyigX0/BnovxzEsWmp7Ivo0y+KLLlEkrlMiqlFHzpIaT6ek7UkI/Hka2O9rmSTaIUyaFUSCB4GaFQU0Vgdgh44Bn0oZTMoDI7FaoWERCBOURASlovg1HKIO+LYzjsg8fk91iIRb0oFiu9XD0/83iDCAeqVFr76V6NCm5pjEWwVq3D6/X2U4jyisCCIuAJxTGENJILbVG4oEZZnRWBzgQWrpJWySFO96XXuPBCQ7QQTWwVq1cqqPn9Yz4+HkgUYdyE4Po2FQohQYtQJBhC0vj86mVkYiEEg0H+CyGWLdMtWEUmHEC82ByIOnIRrpJpRWrVXE4iGM5QiamjnI7SrWghFIognq04eeo5RP1RjHomCxjyeRBKj5qmqmnWm6qwkhqKqajbBipU8VxrJV5KBBBOZZEIW4ikTd7RVM1GEKBrtc3YNeZ8+x9Ur6hgjT/NtmdiCFoW67YQCA4hZwqrF5AMR2l5KyEZCSGSehWFRNhmU0yGEYqmbXdplRa1iM2N/yJJFJpennwMAebJDIVgxej6raQQCieQHoogHGJenivSwhkLs//GHZ2puA0jB5Zv8Zg5HkmXRnmPb7r+FoFZIFAtUFYEfJw7fvu+LY31ZHZogR+xhIV8KjfBvJxortMDYFE2pROIcc5Z/H3InoxOqubjCHOehSi/QmEjAztUrUMiIAKXlkBjhhN719i6desM19Jv8YcbO9cNNEzbWv/WbG+88mHnck4/uakxePdLnU82jjYeXT/QGLx9d+Ooe/3RR9c31m1/pWH/+eErje3rNjYePdpomOPrzS92eqVx91Ws/6q7+ZuT3tm1qbHRnH/nycamdc32nG7sv3OwMbBpd+M0/3ty05oGi3bS4Z2NdezDstv3OXXx//tuX9fYftAUv72xbv2jjcN2ztM8vqaxZfdp+6+D961pLGP5+99xinllO8/tY2vf2d3Ysm5LY7d73Dnr/v802zQ42tbTr9zXWD94e2O/KdKcW3dfw1TbeGdXY+OaOxsvuSwOs8+DdzcbbFhtbOxqK//gfesam5502tX4cH/jzjWbGk+654+S+5o79zt9239n46plGxs7X3HzHn20sX5gXeO+g6Yi9m/LVY2B9Tsbzp/sx+Cdjf3mlGlPs20fvsTyb2+YrirNXwI7duxoPP3003Ozg0efbGxc1iZ3OH+vun0/7+DO6cPdWxpX3fkST1KOrFnX2OlMaFuWrGv+MeFcP9i4b80yyhz3IsqL9WvceXp6X+P2dZwjbsWn993eGNxiZIySCIhAPwSMfmPrEX/1wNh/K661j083LUxLWiWP3Btnx2rHb+UmWUnajs4JUgDRRBR+OwtXtTT/hKJB1zVKq06wgkKxTotdELWCu+GgUkTJtwmbPCU4XtO6fV0w5AcDtFAJRhC0y/MiPBR2rXhernhH3ay1YgG1TZvg508naL+EQtmiG5L2vWIRniitYnaLWUbUQinv5DJHrKEE6L0dTfUqskNJeIbTiLYfb8uCahZRs3GAVrLIcI1WsDTYnLHJF0OOFq2QiyvAvKiNrt7H5R77ZzmPIi2FEbd+Py1vfrJoLvCN2yceHK3QEwgjapmKvLTY+WkBjMD5k1Y8XxUVY6GgpdRTq7i/B5HMDbfa1rUtOikCM0CgVsyi+MHYgt/N89ikdQURj3uRTuYvsAR3neseyqaIIwU4SWDRnm7PRs61UiCCkDudvGHOnXKxNdcmbY4yiIAIzAqBhamk0c3gWzaO7wAf8hMpJ3bWid2hRhHwtXSHGqo1ulDbyvL6vajVqDFQqQgaRYRFGQWrHkrQ5VpFvmi0CSpYFUfBqjOvp92PaNrrKj1+uibqrqJXKlQQGhrmrku6Xissokxlzc86mLdWrY2J9fKYBlJZavZibBwYnbF0USY9w0g3NaROt19z40CJbc2nMdSmMI1m98BbLyGTZBwNlcRIIk+1tcfENtd9dAM1s5OBl21uXm+YuBicHJ7RvB6e6Bjb5o0gNcw9H3R1DtFt7OXO1Fb5PTZL2UTgYhHw+vwX3n/cMe7voQJ/NI5gITVm4425rPtcbw9JcGaPLQNqdS4YKTvCYfcfd1fzWBcp10MLlUUEROBiE1igShof3KktuKpFcxk2JlOITaCkebwe1OvjxFe1jJJtqukjeUywvbF21Wksq8AKMx4kYnFRy1dZGMuaq2CZEsfX16rdohJWoTLGuLd8mbFlfH1FmIpegYpezVjPGJM1JSWEClC9TSHqo1djs1b4Og4qn+VADIlUhspa0wo45RKneaEHAcawmd2j3gxj02ITxfVMsxpdLgK9EAhxIXTnIAaaeZetw33puGvxnqQALjgS0QqSqYsTV+kNM+Yzn0fe/CuUUK3SAj5JE3RaBERgdgksTCWNjP18WJePvoR9u/fhpcN0R8atsVaatnHwmNUvNw/YbgI3lfkKjui4oHvnFK1qnhoqbZlrVOYcK0/TXUnLV8nPoF1uHODGgABdesVCgb87LlJjMTLWtFaqVszC102Oopc3Lk4PA+6pQAYjQRrRiigYN2vI0TS9tJzZ1js31WmlgtfdnTpasvubB75IGikfFdX+3qlxQUm1QhZlCv8klU8/28Ce9L46Z35PddRyxg6gxjZPSelstowKbbZYhccfoaszi3Ax5WxkuKDlOiACs0HAh0iGr/F5ZT9279uPg2YjjRPX0FPlFl9h4+MmpXybfOlvrjenPBeenF8tsVIvIp0Y3VzUU2OUSQREYMYJLFglzZD1+rnLkC65UGASNSAYRZirzCSVIDvVKNAydYSpiFyYvLSQeVHIuatdCr9s0YeQK4h9oTDqRsh6jYJlGhFiXBd3PmZHFSy+2RWBYg4Fxy+BfCZPBbEpTqmU8cJiMoc6rWZGJfMwfi1A5SNNdyk3rNrJz11btVxT6PL9ZNkS2+WevLDRduxWOJWCP023Z6lTht6O2W5VvkfNeYYwzi3FtrPprYcBf2vXP8eUatzB3HXbfABVGatTMSx6q7pzLr4sd3jY8GOiT9TDflKvVRKBS0jAw/nJeMoILbvOmqr3tjDmMxEqIkXXfTP1PdfNhZxXFl8t1Cymlk8hxW2mmhq9D4VyisBsEFjQSlrvgE3AeRz14aD9xQGLr36oJbJcAXcuITCUQbwad7a20/VXi2cw5HfzBviKCFp3aowtc+QzLWrhuh3w31SwwNivZLzG97IxXiQyhILZREB3a1PR8fHaGt0TgeYFrqJX9FHwNpsUHEYmUkSMrwAJUTFM+5JITbgjwL2IimMqFeDrQ/j1gFGtqnMnJzjqCQ9j2MsNBXwIRSIJVBiQHKg3V+x+bqjwIE1lMcSvFVyQaBlMsv4sLYMhWhUjOYuB0uHpPTiCCST9GUTIMsTXgZQZwxeZRCe/oF06IAJzhoCHr9OJwnPs3dGFz5TmOl+Hk/JTRnCTDedaKOXFMBdo/eqMcwaLGiIC85TAIrM9dCb7tmjRInCLKvbs2TOT1ahsERABEbAJPPTQQ1izZg1uu+02EREBERCBGSWwbds27N27F3wFx9h6DlHnOXnYvINjWvXLkjYtfLpYBERABERABERABGaGgJS0meGqUkVABERABERABERgWgSkpE0Lny4WAREQAREQAREQgZkhICVtZriqVBEQgUtE4E/+5E9gYmGVREAEROByJ7D4cu+A2i8CIiAC7QS+973vCYgIiIAIzAsCsqTNi2FUJ0RABERABERABOYbASlp821E1R8REAEREAEREIF5QUBK2rwYRnVCBERABERABERgvhGQkjbfRlT9EQEREAEREAERmBcEpKTNi2FUJ0RABERABERABOYbASlp821E1R8REAEREAEREIF5QUBK2rwYRnVCBERABERABERgvhGQkjbfRlT9EQEREAEREAERmBcEpKTNi2FUJ0RABERABERABOYbASlp821E1R8REAEREAEREIF5QUBK2rwYRnVCBERABERABERgvhGQkjbfRlT9EQEREAEREAERmBcEpKTNi2FUJ0RABERABERABOYbASlp821E1R8REAEREAEREIF5QUBK2rwYRnVCBERABERABERgvhFYPBsdevnllxEKhWajKtUhAiIgAiIgAiIgArNC4NChQzNaz6woae+99x6MoqYkAiIgAiIgAiIgAiLQG4EZV9IajUbXlixatAhYcS2wdlvXfDopAiIgAiIgAiIgAguJgGLSFtJoq68iIAIiIAIiIAKXDQEpaZfNUKmhIiACIiACIiACC4mAlLSFNNrqqwiIgAiIgAiIwGVDYMZj0noi8dF/ArVKT1mVSQREQAREQAREQATmNAGj11yEtIiB/d0j+y9CJd2KsDcOKImACIiACIiACIjAPCMwXRXrklvSfvCDH8yzIVF3REAEREAEREAERGD6BC65JW36XVAJIiACIiACIiACIjD/CGjjwPwbU/VIBERABERABERgHhCQkjYPBlFdEAEREAEREAERmH8EpKTNvzFVj0RABERABERABOYBASlp82AQ1QUREAEREAEREIH5R0BK2vwbU/VIBERABERABERgHhCQkjYPBlFdEAEREAEREAERmH8EpKTNvzFVj0RABERABERABOYBASlp82AQ1QUREAEREAEREIH5R0BK2vwbU/VIBERABERABERgHhCYZSWtjERgETweL7xe88+HaK6KbNiLcKY2KzjrlTziIR+80RzqPdZYTlrwDxW65i4lAgilq13zXIyTtUyYbSkC9TxiviBSlYtR6gyWMY121rKmr92599XyUhyBQBykpyQCIiACIiACc57A7H67s1ZEqboJmWoeUW8bm0gN0VlBVUY6nkK5WkMgYsHTU501lAoVBGJWl9xVlPjktyK+LnkuzqlSoQwrHACo6GaqYbvQeqWAYi2IkNVDj+pl5NnWcIhlzEbyhFvt7KW6ajGPqj8MiyhNXwOhbtx7KbEtj8WxL/d5jbKLgAiIgAiIwCUiMLuWtFIBJX8I1v/P3vvAuHWd94KfHxwPAygQ7ToQ/aeeG0DBMK4fhsY6nhtA7nD98naY3dSiuw7EuN0nFu1abOvnYdw2orHelJvNg5j0NWJco2Lz3kJM240YxKgYNYAYNI0YRIBo14CuW8diEQG69rrWDepIN7DQuXYM3P1959xLcjhDDmc0M5oZfge2hvfPOec7v3POd7/zfd85X6+AtqkNj1OuVqZkJEYJ0xixZouabQMC0BCiPbxjQ5O2jvLE8sS1qdWOkbmIFo8a+SxV3REENBTqVHOUb4yqQ1yeio2726RipkJt1RSbLCs6HPeNI0RKFgQEAUFAEBAEbjgCmyqktVsW1E0mLdLhNHNkpGvExk7PKlM6Eae4maEMTJKZOoQJNpcZaeKf0K1QAfmViU/dT1I2l4S2BRoSfNTr+RSZSZRvJChb45eWpogSqBI0siLJtsjyEpQwuCyP2tUsJU2TEvE4pUotbTLld8ihaobrjlOmukzdTp1yEAyNuEnZtBmY8VyYeg0KLXrNnEHpmkKCrEoGwhjq5XrKwI2Ta1HLiZMJAK0C2lA5S7WsSbnveNQqpugPv/QIRZMVUKJepnomHpSns7craUoWWuQ28pT+4v9ByWiIK4S3CnDMNslrl9DWJGWYxijoCdviy459IQAAIABJREFUNKmUTqLtMP3G01QJNFKeVUFfJUGTAe0kBCxdVedfTaeNJtUoHUXb8xlgH1N9rukMXnVhhka/V902lTM5qgPTph2hNnDQ71fRw5xG6Ge3SQV0cDyB8dHB2qEKpOhCAGW7mlHYJtIZSqG9SyzVA9oLgKiaRXsxTmPxDIXDzG2VgYOp7icYBx6vA9u8/Djy6hmKJtKUTQHjGMZRIY8y42SAPh4CXj09tH87oMsPQUAQEAQEgZ2BgL9p6ap/Yv+ETxO7/d27+f89/sGTC/6FIzP+zJELvr9wzj80NeMfubDg+1dP+HMTM/7RSyDuwhF/euaozz/9q8f9uclD/jn+zff3zPnH+H2ky8dm/amDp/2r+L1wYr+/e/8JXz9Rj7vp3CF/cvqIjxpHSgsnUdbscf8yv33+sD89fdg/zwVfOurP7Dnon8bPqyfm/N2gkcvk3xN4n+noJm77lL//BJdyyT86M+HPneA3zviHJud89RO5j8xw+/Hz6mn/8IGj/jmu58xBf89cUB7Trn5f9Y/PTfnzDMTlY/4s08RFnMHzECvgMxPe52cqnfcPT8/6x5iMhZP+/t1z/nEF2Dl/nu9fWtqW3UF9Jw+E9DNJe/xZVQhonp70D50BoQtn/CMHA5p1Zfi3h85z8/7kJPDi+hi7yXndj513db9NHjyj7yx5X/f7yv181Wda546jMb1YL5z2D0we8E8zpjx2plE/k30a+E7Na/w6tOgydH/1tpezTvvTAH4BbTsxt8c/wAMAbT80pfFjjOen0DfcIUvaELR5wDg6f3jKn1RjeAFt2O1PHT6v6jk+t9s/gLmycv92GiA/BAFBQBAQBHYAApuoSWtT0zLocMsl1+X/oXlKQ2PUcqBci5PXrFDDyFEuzrYuqCFiCWKLpNtqkgvtGX6ykxK1oR1hq6K6ny4E7ztUr7XIaZUok0pRqtiiaMxY1ufMhjbP69fmDZG3raZFMWiPYninVa1TJJsj7foFGrH5ge+z75SZy2oNId82YovrhpaoaiUpn+G3PfI8No1GoRSCBi4amH97tGRQp1A65VENZsx0vkHReILwNjHtUdAehd6u2YZGDUB4wMGOQwvGbUCbo8DVRR2NUpWMYkFh1UlOoBVkMmDyjUWZFmjRakVqmkVoJpe2JYLNHRGvSdWGQ+1yllIptAPm0liMKYLZGBXUsmkqNA3KV/NkLrK6dul0LIsi6RxhjwgqtMlFmxZpVHG71wdNv5/veZ/HwAj9zFi3TGCNxjDWqEVh3Q5M7aDPqlbRjyGtHkUwphbRMrC9FlWrEWgDTfRvlDLABHsb9NiN50hVyWPXi0ILiWYOaPPy4wi+j5ZHqWwKJTtkO9BiZthvEriz9pQH3Ur9y9VLEgQEAUFAENgxCGyekGZj0wDMhrDe9CQW3LSwYfMHDV98/oaHwhi/qhzlg0xsLo3B1MhygNXi+6EIwr5aJpWaTWo0GhBg8JGrmMt0EoRCVd4i0WXxe04VJtccabctm1oQKlmIxCdXfUQT2u4JIbFFtvq428p3KmEqyjt0MY2dZDXIgplT1crCmAsBFEWygOWEAiMLERC2EvjIN2HCzVsJypcgdBqwEKv2BwJtEjfawDIGkxsqYUwMtEfVF4Pg5UEAblep1M5SMa1pCunwWMhVdXBiQQLvQnArV4jyxSTK0G1JqrYwVpZy3I+gvjYE6Dqc+hsN0ImNF7U018i7ciGIlQxqQqDI1JywKv23QyeXZXdwZ2Gz09edHMDaigQ4hu9zu1mW1e/HYExdsZ8ZRwj4StZXgq/GmgUmlihjEGEtCxtHgn5sYzwYkHYVfmEa1F6nhb6DOdPofZnlPwjPKEOhjXHeIpg9jUFtHjSOtECrhiZM8i0bwiU3nzfb8HjhOlfo38VUyZUgIAgIAoLAdkdg04Q0D9sflYDQ+zVkwQ2aJJYJPAcaCJXgS1WB9kh99CA0tFmBwC9AiwEhQDv88we9x4GeBRNkj6Bsr10mM5Ym5dq1JOFDiHyJ0PHeg5YCAl1Ys3qdtVs266JASbNEFTtF2SQTzdoN/FH021StWErroXzcmEb1Sp0qTfgSpWL8Uie5DtehG27XKtQMBCUWNKPwz+I2N9FmLbhYVG9CSwP/OsOtUxUaKqVFgYDCmkj+7QJLFwIHa5ZaLeiKeCskJ2i9WAtThxYtVsgt1g7hMQs7UUiHmhIIdDH2jdJaNK3k021RZUEjVWnCt493rELz6WkRBBsPUhSD35uNV1oF7Cgto08yJRxrElFaud7UpRO0qw0PXDMLm9gtu1haV/3Lmy+UMBK0NanUclrYUe+P0s8YCIw151RYw5+LBd+ulo61uFwHiIXGrFKHwKYr5Zs6DWgvZ3TVIAMk8A8LfSlZM6wT6qlUoeHNQkwb1OYB44gFw1CrysIt082N4HkTavqW9K+LPrWVP6ckQUAQEAQEgZ2HwKYdwcHahlgir8yDYWLBzUlktSkzA6foDJzl2xDO2lB65PBRxucnDpVIEQ75Lo6dsPERz/JtD9oKmIDy/JtTBM7zGZdycNouwYyXqtQWH/ERVhiY+wpBPgfO42Y9R3Yj09WkJDIoJwPTnwHhEGazRkWb3CDyZNixPwNNSjRCsUwVpi58RaENa+PdSBbaGztCZqmG88vCCvXfKBziU6UCmWYNH15bmU9Z6xIDvW4ezvwWBJ822lrUbTYTDuWSSWpBQHKBDqynEADx4Ya4wWWHGi6WFnASBzYIJMmowVke0m6MYO6D83+rT4umcYIIV0pTOlanOoCMxWCuqxrYBKC1k6zBtLgytCUBk50JYU+1JZqlTARtiMNMCOf/MnaIGnwbah83l6JEnTdElOHwv7jhHTohbLI2qMiZWBsGYXPJcSEeMI02IezlKVpPqLayfLjo/VH6GVinvRwlkw2UEGLNGsIYJYss9RgQrhNUxoaLVDJGDjRthQTX05PM5dtLOBoki00TWTa/Q9CuYFMD92Mik0X/pyhei8LUnaNaDQW61QFtXn4cuWyux/zgEeBAS4vdB2qusNYxZgITRV5f/6Kf06k2laD9TfXSL78FAUFAEBAEdgQCN7Ff3dZqSYtyRpFMmAiz+su0tci7Lmo8teOylm1rAe+6ylouM3a/xlPklLEDUTl/DU/NXIxyEQhQZS2k8Q7PpI2dsji8d/unEbDmnazZCMy4eb1Q2PKNXl3/bvnmCIGCgCAgCAgCQxHYNHPnUCqcCs4uSykTpdeswkSVDrRXQ3Ntw4cDtEjr1BLefFGPFqg4goAGex1VGiYVYbJk/RJr5VrKbyxQM64TTZtbDLRKMGVnGqgVpudqC35yi3cy6GMszDJ0bDBNVmtESZiVN5fINde2qv5dcy2SURAQBAQBQWCrILBp5s6hDY6ZlMbuwjz8pYoRmJ9qWWXq2XGJHc9hhOuzCq5TM2G6LDVhoiyPJHS0KyVq4+DYLi0bK0CuUyNXKAZmVJw7V4YpN4Gdt2x67rf6RmBGTDoFOOVX4Q+YpUp1uwilq+vfFYCSx4KAICAICALbAIEtaO7cBqgJiYKAICAICAKCgCAgCGwwAlvD3LnBjZTiBQFBQBAQBAQBQUAQ2G4IiJC23XpM6BUEBAFBQBAQBASBsUBAhLSx6GZppCAgCAgCgoAgIAhsNwRESNtuPSb0CgKCgCAgCAgCgsBYICBC2lh0szRSEBAEBAFBQBAQBLYbAiKkbbceE3oFAUFAEBAEBAFBYCwQECFtLLpZGikICAKCgCAgCAgC2w0BEdK2W48JvYKAICAICAKCgCAwFgiIkDYW3SyNFAQEAUFAEBAEBIHthoAIadutx4ReQUAQEAQEAUFAEBgLBERIG4tulkYKAoKAICAICAKCwHZDQIS07dZjQq8gIAgIAoKAICAIjAUCIqSNRTdLIwUBQUAQEAQEAUFguyEgQtp26zGhVxAQBAQBQUAQEATGAgER0saim6WRgoAgIAgIAoKAILDdEBAhbbv1mNArCAgCgoAgIAgIAmOBgAhpY9HN0khBQBAQBAQBQUAQ2G4IiJC23XpM6BUEBAFBQBAQBASBsUBAhLSx6GZppCAgCAgCgoAgIAhsNwRESNtuPSb0CgKCgCAgCAgCgsBYICBC2lh0szRSEBAEBAFBQBAQBLYbAiKkbbceE3oFAUFAEBAEBAFBYCwQuHmjWzk/P0/PPffcRlcj5QsCgoAgIAgIAoKAILDpCDz11FP0ta99bUPq3XAh7aabblKEf+ITn6BbbrllQxohhQoCgoAgIAgIAoKAILCZCLz33nt07tw5CuWcjah7w4U03/cV3S+88ALdeeedG9EGKVMQEAQEAUFAEBAEBIFNReCtt96iu+66i0I5ZyMqF5+0jUBVyhQEBAFBQBAQBAQBQeA6ERAh7ToBlOyCgCAgCAgCgoAgIAhsBAIipG0EqlKmICAICAKCgCAgCAgC14mACGnXCaBkFwQEAUFAEBAEBAFBYCMQECFtI1CVMgUBQUAQEAQEAUFAELhOBERIu04AJbsgIAgIAoKAICAICAIbgYAIaRuBqpQpCAgCgoAgIAgIAoLAdSIgQtp1AijZBQFBQBAQBAQBQUAQ2AgEREjbCFSlTEFAEBAEBAFBQBAQBK4TARHSrhNAyS4ICAKCgCAgCAgC44sAxyfn0FCD/r8eZDY8LNT1ECd5BQFBQBAQBAQBQUAQ2NII3PIhog/etpTEf/tXol/829L7q7gjQtoqwJJXBQFBQBAQBAQBQUAQWITA7R8j2ptaCspr3yZ6+8LS+6u4I+bOVYAlrwoCgoAgIAgIAoKAILBZCIiQtllISz2CgCAgCAgCgoAgIAisAoExNXf+guwfNSn/zVepdfEaubfsImOvQfknUpT7lQ+uAr6NeHWBqvMl+i2LaHcqS+4zH9mISgaUeZnyByr0NYdo+oknyfqNDw94r+f2T/6O4r9zlv65981bIjS5dy/lgGfhftjqaQ3lhuW9c5VsupUMLubdy1QpfZfKlkP2NYOqf/O/UYbvSxIEtigC3uv/RIXnz1L9tbfJeY8octvtlNy3j0pP/HuKT4xI9OVzZGYa9CLtokN//kdU+RUi9+//im794kWiW+6jk3/3GUqPWNSqXjv/txTLv0w/7c90Sxx1fnZj6kRdm9K2/jat4dr5zn+jO7765qKcE7uiFL/3Pio9/R8pdQceLdN3o1blXb5K7h23UgwZvMv/RPlSk+oXXXLv+SS1j32CjFELkve2NQJjqUlzf1Qn89mz9J3XXDCgCEXJpX+2LPrd/F9T+fK27s8bSvzEbVGavC1CE+959Pprr9Izn3+Bqj9bI0nv/ivV/vQ4Gb9eoeJrugz3B9+n/A/epH++EqHEvjjFb1lj2ZJtxyPwj//4j/T+++/f2HZePk/p33uBvvaSQ69fu5miu4h+7jj0nRdeAP85T1gLrTlFsLCcimG+xXaBf218UnNb1afrjGx8lduohptpN+Ny28307jWXXnnpLH3q2R9Re60tuHyB8n/0ZxT7T9+l5ru6kGb1u/QX1tv4Xt1OqX2xTenztZIv+dYXgTEU0t6hGjRoanV4W4JO/80fkfOXn6ZZ/uC/9yZVz15dX4THprSbKfX058g++Qw5X0nQJLf7PZsqP1gjnteQ95RNr0P7ECYPwp/iWbH7qPzHH6fEqJqIsekDaWiIwO///u/TO++8c0MBsU6dpe9dYxJup/k//0NyThbo3GO3K5p+jg959fW1kxd56FFqfwvz7a8+Rcm1FzNizl2U+QLq4vpUnb9Gy7hIj1jWTnztbir9OfO+Ap1/XPcvXbSo+pM1thV5Ky+9TT/vZP8Fudf0gmPiwSTVf+MjIqStEdrtmG0MzZ0ONd/QXbU7cR8llbksRuY9u6h9BT+vMFd9n0q//Tw9A2vC7ONpSrx2lhr3pKj9Bx8l9/w5ylctal58G2ZSVm3vpcKTn6T0HR8g+9v/jT7yPNTf9+yj83/1HymBklp/dpQ+8YKLX3fTsdr/TjmowL0Xv02xz79KP7/tATp38tfIhCo78+x3qe7cTPEHTTJ7BBP68Wkyfq9Fr3P+v0H+X+rJTzfT/i/8IdX/A0y075yn5Kfr9EO05chf/i4VJhfI+s73qdCwyYKKnGIxSiQeoNKT92vh5kcnKPJsm96FoHr08fepdsqj7Fc+2TeGf0HNPy3T/3gKmPS0qe+lJZfRxF5ouSwlYLlXvCXP9Y1fUPvvG5SrtqnleBTZxSbnuMIyE2tTKvNdtIWTR9/4/B/TDx+Mkv0S44jktOgT/9HVJhfWuH39+1R+6U1qO+9T9J67KZlKUvkzv6wYWfNP/0TRP7Hvk1TZ1abSGwbVnvAopcw4d9P8kzFqvfAqtd+LUiqbpsKulyn/dZjBr0Fbl0pR/T9/TJkb3B//A+63qAHtqwtzbjyRAJaBSUNTJf8KAj0IYP7xvOMU20uZX/mA+mk+vJcmX3gb89kFD1nAPP1gZ4zS3gfoyL0OFjYOubvuptzTj1FpZnl7/hKT4M/+gZK/znMmQvsf30d0FmMVqro4xn2tcL82rb57lepfb1BJzRUmC3yP58p+PVd6iB/h578O5pE/DngkTLzurtspAdeH/JNJxSNV+tlPqAQTcO01h9qYZ/G9d1MmG7pGLF+195N/oNxXYTaGe4qH+WdgnqeRpzhzK1rcpWXyYfDiKy9T9bVrFNmboMoXfg316jJ7ebcD03H83jgVmXdPrp2uxdR+ADz2btr9TRawrpHD3xNoT5ekdy5R+asN8Cw2gd8M1nw7mY98EjzrIxR78SRFmS9zpvcu0mf/py/Tn+/9N/oRvkWc3v1BjW668mm6/LWPU+zyT6jwdXybenF8HDiqMbN8/9RjDfrY198GPwfff9ClMgaJdxtcfZ7+JCVealAe3wsbi4rME49R9VPs7oIx8xcNKv7AxvfxfZjrgTvqqOy/A7hL2gwExk+T9q5HbiAExWAq0APtl6n0/0CjdvKPyDr0y4tw/+ELdfoa1MzsT0I/Ow9hqkHfwHUEfgep2DV6Eczw0c83ycJj4967tQbJeZvaapZdpRY+6jpBOLz4C/XThoDHq6TdEGYStED1579L37oILdEuMM1bLsJ/JcjCf+ArZ6qJ7lIrEC7bYLJ6lfU+tUCLSm+A8fLfGIQUqLG8FxuU+urL9L3X3qf4w/dR/Nqb9L1Tddz7CcSennTlVfjMvEovvuEtvs+v/OxVKjVYaI3QXNZUQucoybUukhVgHI0NmMo/aVL6iy/TD9+4mdLsu3YvwUzQos8+CywnYpTLxjWWEESnISw9O/urdBiCmkr4gB164gHQ8wtqfb1Gn32hTS9eg5D18N3A4SJ94/m/puzfL9aivHu2Sb/VYFNprwnMAaZ3Uxblelcc+tZX/ztlX4pRdt/tFIHZ4sUX8EH7Merjfv/8d9HvHsUfS1GeaT17FvT/AxiaJEFgOQTAZwLtB8EPjQV9lTq/38fzvgUM5rB97wPgK2wWtenLX7fWYDLzFJ9Io5Doe9folR98lwovab5jVf+aHlVzBR/hRwyii5grX+W5srBcA0a+t5hH/hNlPw8eCQEkgkVhBjzyh5grIY9knlh+tkbP4KNv74pTbt8usi3tGlEZ5Gry7iXK8/yD4GU8nKJy6nZyX2vTl5+tU7Uvz+sXHYqn7qPELe/TT197mXLfvKTbsYh3Jyiz933Fbx599vvUCnj1qulaghD4UYc376LYMsdmEfh9rVSjz7Egfs8DVH7iPopdeZO+9XyNcj9CP0B4Kj0cjpbb6cAT/4F+53/9JB24R1c2AcHz6GMGFqD/SuUv1ujLCkeDMomIxhHY9rvsLOqfkGaMNQt1pVHuT99o0zNw9anedh9l90boXfDCbzzfpAZwaf1ZlR79Zpvat+yFH+V9ZFyx1ZjJn9djagkEcmPdERg/Ie2998kLBIjIrhEUibcYdOQrObKeNMgFM2myzLIL5rYvPUq1LyRpmrsEg7zGqm0WqJTZ9G2yWKB6500IZvh7y800Ae2cFqigQcIKkyB8JB68G+LP21gJaWbNWrvKMynKBhOSi6YJCF17+cc1st9gweMqNZXgx2Vigl0EY8dfB4Ifm3B3Q7PHwkuz0VbXEw9+kurPPEr1J/bq989CS4T7vclIpencX2aUlq83WY0WfY+xiiWouG/5FX33/fep8dWjFD/wJ9ASWtqcDOyyD966uNDw6hZoCp5M0dGn01T6zMep8Phe2sPP4LPTfufD+MjsDRxjgdPDn6Df/vT/QLkHg2UphLTsZz6K5w7Vzmohdebxx6j2TIYq+1go9KgB5hWKx7rKXbT/6SydLyd7hM0YZR6/n3KPQYhVL0HQe+zjlH08EbzDH1rcZobGf7EsTqWAxRNJOvTIA5S9N7JUsNWVyb9jj0CXzxB4Qmep0vM75EMdqPDRLHzqfso/HJjMrmFT06pxjFDykYco+xsmJZXPJuhQwuLlnrkCPvOfP0OVFM8nnisXV6jnGn0j/8d002zw/2/3+Vst4pEWqXUd88gvfAqarCTNMhkhj7wMXomFI8+1XOFRKj/zacoxv4NrxGBXE7QJ2pujT36aKk9+HBuSgra9x4ssLrybJmGJyH1qH+amvhcKyot5969RtbCPZviVN16lKvPoNdEV1vsmFX/vz8h49L/CaqL50cQ9ccp8tJey7m9jXxJtgTYK2qvcZ5KUU/zdI8tCb9/xEco9HFW8mmCpST/2AP2n//kBSseCb9U991H+oQ9T5PLFDo7Zwmep+l8+TVmW7eCyU+t32enpnw4VoC+//34qPBIIhLdho9dnMPYeuVvXDdcSl97paIMjqDcLWoug59Aj8AfGuJK0OQiMIKVsDiFbtZY9DyehPtbSS9u5ptXQt2GVxLMIfw0wwlewYlWq7Y9CoAJz+BYmW9vBSgN+VS0IOVP4oBss8ECggpUBjIUHeIxS97LggxWVEhrhWAyne04Gr8A6zOdDZN4Lpg2NWRuCGEETqAS/exOUfe9l+ouLMGe++wmKQuWtBT8DH4T3QY8W/KKhY3GgNXyXaV2kZLobu1rvJxNmVBYEO+naRaUK5zJnH4cJVnGN4endK9iAwa9gh9M0BFbe3dkv+HVKmNxLqYswx77QouLz1+jnEJ5VYiF6eDU9Tz2NO2hkrSjRB4K/0ErCbL3oAwcmU9r/ES2McbNUupkiSqjuXkf5GkL1ogQzbPK2l+lbWGE+83v/nazHwaye/LXRd+ctLk2uBIHlEVBCFQS6/vG3/NsD74YC4eKxjUVeOO6xEGq8iMUQ8yykd8HXeK4Eeuply+WNA7GQPvCpjtCJt3t5pB3yyFs8ar/EK9e3g/nsks31vecGGyZuxsLyAkyymMNBjQ7yEnZyL0kTd1Aq8SZMey0qw4HegdCplF8sgPa9HGIXwdzm52HislWekHffkaDqn9+t281C4htroKtTOrR2sJ5wmoCmNAHeXHr6oQGWhw+S+SBcLGCmLH2zSVnwad0Wzr0KwQe8VuMGlxulsQv+4qYDTJm3hWnRN6xzV//QOIW/+8feh7ATOUYT2E3/87MvUGIeiwgsUCu/stKCva8SubwuBMZQSOt+mPUKU+PnwclYCUvwd4j1jMHIri476n1f3w3hg+niPVb/9ghU0JZZu96ERmkXpR9+gOKvsZAGgepnEcKYh4xmKLOkmpihZg9CBq+oehkgv8EaN/Zz+Okbb5MNYYwFv+l90ORcuQghjc2o/0oGzJVa8OMjRN7paAs5v05oN36wMBSae9V9fBB411l/cn7QohozVajAC6llGGd/BjCF/V/6P6n+0JIHy97wfvx9Sn2Rfe120ewjScrAzFt4we5xll022+KbEFj7mXTnhX5hDz5v/biOUIN+5UMfBUP/TTK+iuM/sFPvWzCx1k9B/f+Fxyh/w49sGbkV8uKmIhDwGa4T87V3nIa/lRC1Walnrrz4Qo0+9UJvxUuFncVk6Y0D1ft77yLcTZB6eWTHxHulTZ+Db1Vv8phHQkOj2/82fQNmv2/0vOCFC7XFlYOdXaBcHi4hkOFYU1baBx/a51+mFzuLq/4MS6+X8u4PUvxXelxb1kJXpxoD/sK/pfyFV07vUPWLf02fs94n1rblH4c94BTMw30awRXL6dDb82YwnniDVa8zXG//rFhu3wvxz2SptatB2ectesV6mX73916lyiNpqv/Bx+QIkNWCucb3x09IgxCmtCWY4Ly64uEcgQmxki/T56Chmtj3GLn/JdaBs/fDHsUW6zApRtMRBCDo3KKdT+OB46gNE1nzGlZX8LtI3/thiiawTn0BK9izETiph2ZJLqS3TBb0FjN0dQOOtyYc8b+HMhsQEn6KVVMmcQclrkF4g9N76zWbHJ7kHcEPH4hA8OowvpDWAUKZqqcnRe+BX5bjwvk/MPmNoEnrL2PYdfvsRQhoSHv3UfUPcObPj10qrlZIm8CKPujLjukoZPRo55qFsmUIj9zxUSr9yZPwxbAoD6ffb8H37XPwT4zXHqXUOmOzTPVya9shAD6j3CmwCLsCn1b8MrgN0Ljwb6U571kAqlsbmXiuqPJ5s1EBm40CZ/l1qLN3nnVcSAZtNPpxSMfddBQbqfLaSDGciott7WZCe6kIE2r2lgvUYiFteK5FT5fwbixkWz/Slg3j3o/Bh20NdK2i/s6r79rwOWaNWYTST/LGEA9m3jUIaR16Q7N617zeP66ujw9+kBKfepSsffuo/s3vwsfPpldOvUCZe56k1mdGWbyvBSTJ04vA+PmkTcQoGfh8/RxneTXZ9Me+Y5pzwlwWHfhxj0JVrr7HUDXbrKNmlbNazXWdRCP3ar+0dy++TBVMxgnsIGJTodr1w/5nfKwEGKWp/NE4MTPnv9BwKRMlzBKBGUI95vQhmFGZZjiYVs6CUDiKwmULu5fgfwYhxTr1MnYj9gp+H4TTqi7954EpwwGtarMBTJHKVLtCiiRw4ObDKONaG7uq/qljklgmrwHjAAAgAElEQVQh2+iP+4Qpr988OVJJMDcH2DkQKBHJFthpk60yz4xUxggvnT9N8f/lCEUfrVNz78exW+4BvakB5s/QZDRCKfLKWCGAj9vewIDoYDPQT3gB9guy4P+lFidYaCX2stZ7sxIsBMos9j7Mnnqu1P+v/1v5mUXmz6/gkzY6jex2sIhHvv4jSihftqOU5004nXnJLiFc7v9HuUe1r1vi/+1q5xbVGGoieeHFizL46mm+OzpdS3j35VcpDyf7R7GrvswL3LXQNXr13Tc7WlVui7acKFeZQWmQtrBDb/i9uBbgibW6cv1YjwSt3x/9CUXB+xLffJ/ShzJUelAvPLRf9XrUIWWshMD4adJgksw8fh8Vn8VZaY5Fn/r1i7Tnlmv0U/Vtx66nFKvAl2cWURy7kNx1EWcftanwxb+lOlT6r3C2XifRQKD6HsyQ7J81s4934iBBeEvQq/RD9itjf7REyKBxOCEc0P/iLHZlnWpSBbq9+hLV963wdcPEQ95X8GxiHzRrzAkn7qYUhLcfoi4W/OY6gh9R8uE47WnAgd9qUuZP3yTvJZupoD0P3kem+rVSgonjCWzPxyrvxbNYQf29gdX3+vkiGPfGILRCcMSZQKXT2Lt66s0OQUorBvOk1njCsfmbf0vVXZ+k5BKS76A0dmJ+Gc66r2AXbtbZRa2zepWaeph989YpYYt8DBrFf0a/l776dxCI21povWcvJUfRBKwTGVLM9kIg8cg+msWO6h9Co/7l3/mv2D2HjT6Br+juB/dRXrk7bFbCJpl9UfoajgN65ZvQhLwRJSucK4/sHeqPthoKo3ysUcAj88+epDp8WzWP5J2DKGkCx5HsvZlexFEa1a+egCbd1W4V2EiU3cdHPiyTcNwGH+nzIp+7WMUhwM7LZHU06Nr6sEyuRbeW491KExeL600Ga6FrpUqXe/4h+Kxh9fi9N65R84VzOCoEGyYCBUHH6sG8D3l/ik0A5efPUeyJxNKS7sDGhHu/Ty9iE1n1q98m7563OzhmHmQ+rWw915lw1Alo/flLHnYJ4yBxKCNagRaQfdUkbQ4C46dJA67Rh9LU+tI+2n9vFIICH8mxiyZxpMaRcpZKA3bkqO74pX9P1S9hO3TidvIQoaDmRGgGPhInvtS7Y/DDcPQPVzK3UzoRCDa/BA1buGsTZkmz83H/INTen6aDe5EHO6DKLzGzWipesF9aqAAzHwwEP+qtq1fwg5ZtJkWNpxM0B8bYgrDWhmA4l/o07n1kdOHlDmwRTzG78Og7OCak9rP1G5RRnN9UxlbzPbe8TXWYOU0IhElVfGhehc8X6J/BER4usK5eXN6p1nwiQyfwkZnZBWaPM368eww68GSGqnx23HolOBlXvmACS2xzx3ECrCGN4yN78iv6LLz1qkbK2WEI3HE/Ncppmoej+OQuOJfj+JfdOK9w7rHHwH/uXz9N70iwfQBz7Df1XLntGtV/gN3P4EMHn85irqzf4ot+6f4OjyRYKmo4YmcaprLTPFcUA/sw5b/0GB3edzcZ2FhVPXuNouC9R7+SGSy08vx7Ik5TwND6AQQ07FgvsMDH1ofwmJOVMOjn3VjsTjHv/kqKkmula6U6l31+BxWehvAOvubAx6t6BZsMHtEaVz6SRYlW98M5H8ds8Jh58QeWspIsTbdSHj6xjGMMvsnVl65RDFabw19aukt/ad5R73yAkuCvR1M4Wgp9VcEh8DZ21h98+jfXd8yMSs6YvneTj7SRbZ+fn6fnnnuO/uVf/oXuvPPOjaxKyhYEBAFBQCHw0EMP0alTp+jWW8VvRoaEICAIbAwCb731Ft11111Edz4IrXBqaSWvfRsbnC/Q9YhZY6lJW4qk3BEEBAFBQBAQBAQBQWBrISBC2tbqD6FGEBAEBAFBQBAQBAQBhYAIaTIQBAFBQBAQBAQBQUAQ2IIIiJC2BTtFSBIEBAFBQBAQBAQBQUCENBkDgoAgIAgIAoKAICAIbEEExl5I89z1OE9mC/askCQICAJbBgHPddfl5Kot0yAhZOsi4MlY27qds3rKxltIswqUKbVXj9oac7RyOICwLkLhIvisPCXSteDEc5dapTQlTZOy1Rrl4hkSuNY42CTb1kHAa1I+V1v/qB3X00KnggO1Szg/cWsmp5GnFPiAmW+sm3DbyiNEX23n8992OUOF1tbsV6Fq9QiMsZCGw08hoKVyy5zmPAxHq0LZ8lpmgE0tK0pmYulBtcOq2+nPnJZFETOhTzy3SpRrpqnWalE16ZAVNUng2ukjYOe3z66WyctktlRAas9qkY0IKsZWhN9rUCHvUK7RolY5Nfrh20PbYpMF/ps0x4H/Ilyhs/OF0aHdvYMejq+QBmEL4hZlVsml2o0qOdFVZuIB4+HkaDdB5hqy7qDx1tcUj6yWQwkzru7bzSZFU0l1Ejt/RNyEuTU/Iju3Q6Rl640AtGilmkFZFblj6yQbUTyMZGKdBKB1ble7Qa04NOrrCRn4b9NJUMJYZ1q3aHGeJ0LaFu2aVZM1pkIa4kGWaxTPpbQGp5kjM1OiUjYDFTvicxYaZNVhCk0j3FPchOqYB3ybKhmTUtC+tSt4D2p4l5qUjWepEc4Hu0xmsqLNGjClJlIlqhZSZKYr1EYMTSueJCWOODXKmGmq2v39BaGlkqN0CqFKEnHQ0QzMgLhfzVEqmYQmDoJetkZhVqtgUrpQplwG9UCoydXBjEpZlGFS3MxRw+mvY5lrBytX5E8lkSeBkEqqcI9q6QRlK2XK4n6eG+nZVM+DeSo6UG/FUqYIt5oiI9fVLrr1DEUj6R5TJWJeJlMqRp3XrlIWGJswZaRyRapZCJGVsKmWTVKqaJFdy+J+jRpNi+JJLbytR71ko43oT0W7maRsLUCwv5964HFbaDv6IgUhMpEqkxX0s7qPcpJoRyKZ72I80jhaBn+5tWMRcGsYN2aWkkqBM4RftGD2TxWonM9QmsdbMkdqiDJP6TVLwkyZjMSgcQ4hQ8zfvAk3Cnfg/NT1ZqjMPAS01OAf10L+BAeRxAxulTCWYR9DCYvTsnwBr6j7PJeg6U6kqNQMcno10J6jcgH8BwIgz5mmVaV8BjwD/CxVtvprWHJtVdKKXzrtMnhYgZoQcrNwe+jSjiyD5jKT1iwh/rKmzUxmqBJO2naL2kY3fJ9dD8ypwNpA+SEvbpfA//MVKuUyimYz1+iYqQfxg4F13iDewn7WIqMtGVrb9waHhdrI9NRTT3HYKR9hoTaymtWVffWkf2DqkH9mQWe7cGTa3z1zxD/P11dP+HMTu/3Zo+d9dXl8zp88dCYo/4x/aOqAfzrI51844s/MHfMvB08XTh7wp+bPqaurx2f9ickD/vHzV9X15WOz/syRC3hwxp+fnfWPqMoWp6unD/pTKO8S314AjXtQF35eRbnTB07o+6jt2Owe/+AZfumqf3xuN+o8g1/Igvx7dk/7h8+pK//0wT3+/pNL6+Gc3XTBPzoz7R86o+lkLGaOXsLjc/6hyd3+zOHT/iVVxIJ/5tCMP3fsgsKFcdq/Z79/AtkWTu73JzVBeHDJPzYz4dPEnHqmcp455M8Aw4UFtH3mgH8iAOzS0Rl/YuqwD1SQzvnzU7o8oIU2zvgM1/rUe94/PDvnH7sQYHH+sD89fUTV299PihROl4/7c9MH/JOKVqZnyldde+mYPzeDsRO07fzhaX/q8HmVZfRxpF6XtIEI7Nu3z79y5coG1jBK0TxuMB/1AOcBMpBfXD6GuQB+ocfbVf/kgT3+NE+Ay8f82enDvh5hvn/+8BT46W7/4Omgfp6HzLuGzE+udxp84dCJYO76mo8xa7h0fL8/c/Bkh4d1WzWIL5z3j8zMgMfoucRze3LmqOZN5+b9yT1z/nHFMC6Br0z4e/Yf188uHfVn0I4QisHoMd+aCvgb3uqnfWHwXFZzczbg48jKPDfk3fxb4cmJ5z8w0+yA+dpkwPMCfgpepaa34nHoE27OMH4woM4bxVu4PSvzfQ2F/Ht9CLBcw/INwkL59KtfWPr/7R9Tz68n3bx9xcu1U+61atRKILCwWt26MLm5lC5hJcvXvASJZagIXzV1if+jBq84kWyLrFiSiiofcsIk5yWKnUDJVrMN7Y/2cbOaNrRudcomWGcPTVjToXjWgzauSF6xToUlzlYuNSpNMgsVbeKLQHtVNaB5QxDisg1H+tCnJUZMTlNpdaA1a4OeWlJpBDk8byRVpKKpr8iLUiwaEKtJXvovVrqVCDRugW3BSJephJU62XX4hOWoUoKWjHO5dSq3TCpW4tpEApOvEdUBgSORoD5+D1qmSiRJs6hWK54cqoH+TDlJXh2B7dMlKgdwGoZBMWi1VPkOsIUPWk4VxaaJODYOrE+91IS2M16AJjPAAoGloxFgheIX9xMTopNdw2o+VaG0ojUKLV+ZWLFnlSoEZ5mOKcaI4wW1Wl/FOOpWI792MgJug+oO5iePYx4hA/kF8webkuAL4XgzjChFIhih0ShFwZPUXIKvVrlu0NyME8wt6PcrGI/5GiUwP9MD5ifX62bKVM4EcxdaJSuGuQbNb6YODTfmZTAlu70xgC94jSJV40VoB/VcimAuxRTnYfZoUTQLLbMRzDNwrzy0+oaiHW+AZ/D0Hp7a0L4Z0GLpt/pp95rlwXMZWETzrY4faxTYKQxBnXKrCH1bwAuaTeCqyOR/ImQoHg8fuDYsJwE/RWaKRGOEWOjQ8A/mB8vXeaN4i0u2jfYw0ZJ2BAJjae606i2KwxyohzGYQjsBNX0wqNmh1oTpsHNpQ6VvqM72YLL0YG4MGVqbhbLAnwpSBhxTI2AunLGNyY6dRB2nCmY8bLnIQV6AUNS53zOGvBY1LJgQO46tMTJTcKh3mtRw05TSJCCDTW3b0L4VSoUP02jA+WymB6r+brtgMg2YXU9Ni36yH1gEecLiI3GYCsBklU9YMk1hdq9ZpzZMtGaY2wUdYMJxrhvcLqL06y7Vyyxo5nE/ULlbZapC2MsaHjXrNszB8U79tg3BFSZDppexdWBCVU/bzY5p4vrrBbOsN+HvVqQUmy7ZlAxBsR2Pox/7+6nTOGrCTsz46wSmB3NtPNqGGTYKM0z3k2ZbNhkwi3CfjzqOOgDIjx2NgNdsUBvmwHAUDeMXPHYyWkIDJh4+tJ7mOywoBCKZA9NpO12knMEyD+YbTIHlRgJmfIhJQ+Yn12t2+J0WFh0IfPlSlMq1nJ5zfT0xiC8w70zAbSBgj2BHbXIMmAxZEMJCtDNneKHlgpcE0535iYf53Z05A7rexWINiKm1LVI/7cPmcqOBzVlYVHbmebFFhuLPdiD4BVTjj10vUyEHn+R0ivL1YEMXL8IN8NqgbmL+CneXBDu2DOIHA+u8UbyF68X3IcB9AMpyexshMIZCGoSpdoTiSrrg+YuJGIWvQjB/eTUYg6+FvsSAVz5T+qpXU8YTv2XFuruFeh1TwWhaHhhNyJFYSxSJQ+ByKaZWbMskCCYW+0wEdIRvsPDCu7A6cy6oh3mPWmV2GB/8TCz2MwnaxcwO2rjwcpkacQv+KGCsoeN+7zv6gxJ+XphZso9Y9xocmWzFwJCgSWMhzbNrVHGyVIBEiUUsPiTQDpZb8N9LYwXNzAOCZ7chSoOZMDUeXJ8BgY2Tiz5AQ9Sq+/rrRT+BaRUbEHYbDfV/s2WTAz+6SH8/dQBA/+GjiWGwOAXvByTjmQPMIxrzVYyjvlLlcociYLfb0BQH2qth/IL5Q49gojTkFvyh1HTA3MK/HuZ9uRKhPHxQWUPEchsLbU4mrxaUg+dJH59CWWquGeBg4ElYjy2TBvGFpTvUuV5D8UstHHRYBPiDFnB08VaLrQydyb9MncEtxVcC391+zPh62FymHNXhbBfO81bboUYGWLnQHDK+zGqwGC7Bn7QITX2mUKJKBZq+KHxkgYeHvG7PIjzceR5B7oH8YGCdfd+ATouHlIVvxnXzFnxHWrHuwn0w0PJkuyAwlkJa24l1dvnwCs/BxMQcRWLmBMEhlK4WfZRZUwYhI5Sigt2aHUEMO5J4Y4Diq4pBhYwGpSpBC6bDQpzqxeXPS/LA0B1wzpCOdgPmRhf8xbYpEtd3FYU9K2ZmtokO4wNjt6E5C/lgIEQNZ4ttsuwuFuS0qAazSyh8dI8L0Sv7mMFik068ojbCFTWbFRi7So1icBo2cKVMoSy0edCiMSgeNly4bCINCoAWoNbCh0gRyNhCLlNmWm0m0O1ah3rx8bAgHHYhBOMuw6EZ2Pb3U0AZyAGtEdAafMBcq06NNn8VsbEhxhq4ILlNZc7C/gqleRxtHHVqkR87HAEbQoI2o6GhQ/gF8wcL48oI8cCHthkLNToRZZp3GmVqmQVKY4qwkOZB2KjWsIEgy7mGzJP+epWgA21wEQ72NjZLqU1R/WkQX7Cp7WBedCaADc1yBBuOcAO8o0Xgf8H8djCho53FrhYURzl+qA1hLgZJT029JbQPmcus9erFEPhU4G7icDmYm6Hg59axmSteploeGk40JGJ3ebUWJJlZqcqRzdYL2EH8YGidi78BQaEbzltsfDeg6u+OpU7F8mO7IjCGQhqfxhxTmh5Oi1egvMrpMRGy2Q3aIiVHsAbLho+YEXS165ATbqGBAFIp1DrnfWltXMBo8Lra7o5lcSwNJtvLGKFpcrATp5NC3xMIN3nsMHVDIQHMXr3Fq2nsLk0X2NzQx/gUw+hq4pjZ8Rb7FRPXqQuHqSBHVWgZFXPEdvWu1ktVTo7Nkg2SXaVCEz5zoXmGNWnYsVpqpaBFY2Dh7wEhrQkmaUCL1uHpaK9uLrR+xQLV4RejZF6FLTRXBhceaNw6KsXrr9fDrtS2zWVD6MXur2KdhUjdL50PArcPmk4NBWMS9Av3ba6EtXTQGRCaFeNnEwj81LxcTpmARx5Hmgz5d8cjwGMI8yD0DVqBX7gwtdlqyGGnM3hJHLs89bzhBRD8UisuZaFF41HIZbq1EjVT2KEYDMuB87NPo9Wd2wnwGINqnUVjz/hnMpbjC+q+TfBSUMnBDvhqLI/dlzyF+xcp0NAHbiKdOg2db/C/2q+zs0jup52rHzCXuUwPGPJairGwyqANi0LGsA0TQzjPHcxfZv4KNmjYivDn0+c0ai2hdlfhh9pEqrSZCovl+cGgOhd9A3r5/JCyYOO+Tt6C3fc1omwGHSJpxyAwhkIapif7eahZyuYq9kcIJDaYrKxe06ea3IHpE870RrQBhpSmShtZoVLOmtgeDv+1TK5KLpxiE8oRgH0z2J9Es1j+mIN/abMezI/5fAyMUa/w+PgMs2CpwRQBw827JX1ERLZJGRyAmQSNMRwNknNg0oCDfQqbAlww8BILMGqVCU2UobIHps/QX47rZBq4XRCW4IFvwnl/acIRG4UENbIm6sVW+ghWmTnQrUyvgSlTZcJquVgkgvO/CVNBMm/hCIty13eDhTQ0MgKBRbMHdsqFj5aXpnzoYxeB70fWpSLqSWcK1IpCcwC/MIOLB3NlR2Ylly3SXq5DvVjdF0oxqsK/jGlPlyPYaJFDvYv7yWvkcGRABawZycCJ3WkbtKOt6BcbjBxKUMKXk0qpFjSDvMUf2/tjJazI+cHo48itpaGBXF6bylVL2ikI8GBmLVjQnoH8AnMV/CKT9ijPcwtHbzRS8FtllZlKETiugy8ZecoELIU1aVYrAi2anm3D5idrtCIdjRaK47kdaPljmIfpNmvTwCV6xz/sAcvyBZ5L5TiO5mF+gWNs6kmqYkMTk2Urs2e4MNWmz8B7QdXZMX3aOELE6Dm2KGil/qP9OsO15RLaB85l5DYxN02em5qXFcAzazhOQ/PfrnUkngaPwhFBZioDc2dLuZ/EeeUdLExDuTI0kSrl/iB+MLDOxbyll88PLGsdeAvOb6KaUYD/7yJQ5WKbI3ATbw3dyDbMz8/Tc889R9iqSnfeeedGVjVi2RYVTIREaYFZjJhjW7/m4uyiZAtnvZU7myG2dXu2NfE25NwMeRX0R/h93dbt2brEP/TQQ3Tq1Cm69dZbbxiRvPhp5uEDxbLCwNTELmbwIwvzNBToBr67/R+wwFKIw30gGy5it3+btkYLcPZnNoXxhsOTh463rUHtTqHirbfeorvuuotwBAfR3tTSZr32baK3L/AZHEufjXjn34343g56LYpjKaC+3kEtGtYUF9rAeKkoAtowkDbrGfx2bBxwrJRvknY8ArEY+46t0EylRe66Kazw9jZ/DJ8yL4tDw0VAW/eOhLtJFf6/eRHQ1h3aG13gGJ6TZuD0/OqNxn3T6o9CKChtWm1S0VAEcP4eXGUkjQkCiRLONVuhrXpHYWFMHL3jiHssK5QVhsTaHscQSaK2tqySa2sjMIZC2tbuEKFOEBAExgeBSAbnD2bGp73SUkFAEFgdAmNo7lwdQPK2ICAICAKCgCAgCAgCNwIBEdJuBOpSpyAgCAgCgoAgIAgIAisgIELaCgDJY0Fg6yPgkdt73t7WJ1goFAQEAUFAEBgBARHSRgBp3V+x8oh/V8MJapJGR6CNc5CSCDulczjVJEVTwzFs4Xy4TH2l7XWjUzD4TdCGs9NC2ga/t1FPGpTLDMdio2qWctcTARwPFL+JbkIA8PaiYnE+IyIQ1NezqiFlcSilxLLnKvZmWjwfFxeHo2YwV1csYggN8kgQEAQ0AiKk3YCREMaE43MSJY2IAB9y6wbx95AlBiG3jYNhB2O4NM7giDWt/jWmDQeAduPurb6I687BJ9pfdyFSwI1HYJKmCAdWNzZjcRG0tl2hLJ9oG6RkxcKZW8aNh0IoEAQEARIhbdMHAceEWz6o+aaTsp0qVKeWB7FRme5oDOfdRQa3QEVk0IGTB7+0Tk9U+JpurNZ1KnV1xXiOaGZXh9gWfTtG2XwSIdWqmyd0I9JKS8ekUphEED0FR7xJEgQEgS2AwPgKaU6DCpkUggMjNmciQ1Wbe8NDyBOERKmUKYv7eV7NIlZcHfEnOQyKiZAjaawyeY3rVhHENtddfboINRKNwCTRWQDDHIAQSFWoN7x2FeGjIDCYJqVyRapZBsri6tpUy6Fc3E8gRmiuERpAIchVc6CN60S+bE2HKwroQ1SmIFk4vDBNNVUnQlTFcQ4X5zNxZg6KchFlIJtCGxGEM5Eq4yBJnc1B/MpMikPQICwVhzcKH4TFqr+goZJDrN4UJRHuKlloBkLAINoQxxiniacLZZjeUgqrXB0xOUtZlAGMTYS7GUXVs2y/IOwMYpFGO4GXOTxXGFpmAJ0ctDoUnHDQY8ZMo49bOEgWoWw6dLCZMkLxIDQXt5r70cQp8Xzc8bA+6B0jKk4fOpRFRhfhdcwerDuQcl9jHKWAu+rrekDEgPHFoaYq6P98GeGAEMg+wWO0haDROZTB/Qnzph0WzsN0ExUvnTbJj5URQL9XEXnAwIIiGkds20an1wbkRcg6hCLLuWUqd+b54lc9xPbNYU4xT0omEXotHEsYEXXwkzhM7ymEP6ogxq+JWL8qKTo0r+EQc9laW/Exu5rF3G6QXc8jBFqOuKgWTltOa6aCNxzwPx5/4BXgA2a2GsTHVIVCm808AvPbSHR4o66w++9genvebOUpjnBYIQcEM6F4JEEh+VxXPYN4o8xyXZyqjzBvTFMijnoBVDj8meelmU7wrUwJ/A88sR5U4zQQ6xS8VuGGtgICnYAn8+cE8+FkfjQ+1d9IuRYENgoBDgu1kempp57ieAg+wkJtZDWrLPuCf3Rm2j905qrKd+HItD9z9BJ+nfMPTe72Zw6f9i8t8JMF/8yhGX/u2AX8Qrp6wt+/Z79/AtkWTu73Jw+e4btIl/xjMxM+TcypZ5wWzhzyZw6d8RcWzvjzMwf8E5eDN4/O+BNTh/0LnOvorD9zhH/h/dMH/MkDp9XvqycP+NMHTqBUTpf9Y7N7fF3VOX9+6oB/UhHDj475szNH9XsXjvjTu9GmEwGtl4/7c9N4V9XLZUz58+fw89Ixf272iH8+KOPysVl/EnT2p6unD/pTc8d02Qsn/QN7DvhM3WDarvrH53b7U/NnfIZgAfn3gJ7D59SVf/rgHn9/h/D+2sLrQf2y4J88MOUfYvo5cVtBGzdtEJ3cLoXtVeA/O+sfUQ2+4B+ZnvWPBX2xcAY0YmxOzp8PCkY/zs35xy4Na+fSMXJiv8Z24fxRYDvvB8MqKJP/LPjn5jHGjpzX4whjY3LuOHAaPL4AoH9g96R/QHegf/7wFMbNQf+0op3HQbcdahxMH1FjSpJGYN++ff6VK1duMBzc75OK/3X+n5j2jw7sqPP+YfQrs6LLx+fAD06queRj5h0E3zmpfl/GPJvBXA4nMOb5JMYFLq+eAE+aO6Hmhb9wzj88PRHwNeY1M/40GIDKhWfzmAeK5XHiud7DA87NT/n7T+jyrzKf238ioIPr3oMxGc6lCcydoMzLJ/39k3P+cVX5JfDXsPzB9Aa16z883ifBYzrNmvUnaMKf1QXiHcy7ad3OM4cm/VmepJyucr7ZoF7Gj+eiLuQy8NizG3nUe+BhnD/gz8zHwnZxefsDxs18YzrgyboC+VcQGIwAyzVqbt/5oE+/+oWl/9/+MfX8etJ4atIsBDCO5KiQ1Dp9I12mUjrG6hoEWMcKtAQtmVKL1KncMqkIB3S+JJgBjCCkVARBxXl1pxJWb5VIkmbxkr7jUA1esxmsZL16mVrpYic4smHAlICVrIFsbasN04KmIWKWqFFK4q5NVeTNljLBKeQI7A7SlKaE6QsDkePSQ1BzB5o2LsvFbzeD1XdG02rXKuSkEEAZeUE4AqRDwxWHtqtSoWgeGxdUg/AE9Ud0tHl9Q/3rUqPSRPD3rKaBA8NXC2QOo42gNWsnqViEQ78qw4NGoEhFFaGYLzkcV1CpvrP030H9wmVb0OaBfk5uq8kR6xHYeRCdHODYgUbBo0qmSB4C2hdUg6GliEBHpjoJfQQ/nMTcFH7rXvPQj9VYHny+uG8AACAASURBVAGKV+iD3jEC2lptrNwjWI1Ds5qrlSkYVopWlbwmVeqIdJELAlAnMMaKKYoOGV8qKHWioMclFwESzXyJUqo/ObG5N/hpt8mO4Tp8JH+3CALQntZfX0zLu69QbUVtGnoXwc9TrVKg4e8pwmtRw4b2yAzmEkJKpY02tR2OY96mBOI3qnGAOZtJBRMG86TVdCmZMTUf42emTc2WHvfDwIpCu2XVYCVQL8WgTcOsc90gi0EZaAkVJQggn0lwYPW+0obQu+hN0JRG/obKj/kLjMy5BLUbgZaMA8Mb0NCjMrNkQWNo6OzRBAK2u+QwSfANtSgFrZjGJpbOdsPhtRvQrMMiEsyZaCpNiXZLbdCIgv85tqO4QCxTpVomKFvXIP8KAjcUgbEU0uxmEwKEFpQY/QhMYElIZSz0uDAfJIIu8Zo4DRxmMjPsItfGpI5TnCc6BI6I+tq7VC+zQAN1PQtwfMvCxx4f8qzhUbMOZhNKF3hkgxnEA9NYMpcnSDGUKjXJgQAYZ8nQaVLDhUnLCCtFnbZBCVxroQymk+BRu2mRoeymmkGbMDto9uRSE7ZFEyYBnSJkwPQaj4IJgvm1IISm2AzKpsxiC2WEzDx4nRmrBaYZfgiYOaOs6BDaiBkemGjIBG18MOLAWNPThgAHISskJ6im/8+gfiEHzDcKU0TAYK2WDQxB8yA60UuAhuxyjpowz1Q6UlMUQqkWeAgm6IqXpyJLsZ1+tGCuTVFkSDv7x4iizXOpmqtSsloNhOK+llkNavVgQ/iwpLDLYNj4cmBCxcAJ+tohy4KgnQStnEI8NLi4tDCG48GHtB9Vub5xCMQwpyf6qt9NBq+6VkpY9OVzWGSU2fTem3DcCrsEYD7rOZylchtCCl7iU1iiPc5k0VgwQMCjHDeqFnthihrRHmFrCDG8GG3XEM4M/CyThpne7tIT0QtInTuiFg1Lze6D6V1caxTDPYrFlY3bml+ohSrmTht3eGHGPJtZQAS812lUqJDLwhWA2x+UBOHRixndeQDhq7MuBDhus9jheymEy2vhHmObKFQo2Uhhx32Zml7AhxcTJ1eCwA1DYAyFNA+ryuUd93klygJUmFgIivdKFvBzsuMQFvgFMC8W0tjfouJkqQCpSgsA0O6AsabgfxQFe2m1IeB0ZCCsEFsulECaW0bMAlazFUo04AeVrSt/DI/rgHasmwVaJAd+abixmD4XH26ipNJU8U7GGH6HTJm1O1hhmn3jCitN6HqoDobXaDTU/y0swXnL/aKkVq1Lgz4Po01p8pR2ixNW7qAtEUpVvMLFSjm87KMquBzcL1yvA9w1JjbaHbR1AJ1KiIlAaMHyOrbog6g1aS76rVmuwxctSwl8WfjDwn6DVYIvEMt+I/cBy4lMGwQuB5rCzkdxcQuVEAV/l55vpHph8PhiTaAN/xhDF4RNEOEY4BuqTmCtnzJueLcjkOss8u9WQACbAMolmt3dpWVyfwXa0XCeDqcxnoNfVr2k/MR6UwQxYCtNPX8bjSa0aPCt7DCM7ptLBabh9S33VPlo5prYTY0FTblK5WxgVVAv97cD1/23+K0R6TXAsNwm+IiDjQxR9oXF/5EWNW0e4x6EOJ5BmLt5aAKrBB5bhA9uBX65y1QaNIaFsDBxHONmwPca4O2OA+GMH0IbWcJ1NQUfX8zT8rI+uj0FyU9BYBMRGEMhrU2WHVOaKZXAEGpq9eZAsIhAnR9OeA9aLw8f+a4AY9WhdYITt3qDzYRgGK1KjWLKLAhtFWvSWGjzsOpkSQ7Oum2XTaSqJlw3qdaCaQwM1WnWwHxwDxqWUq1IBj68bVy6tg2tiKHfV1lCbR60Ke0e+qBFasLsodoR7GSEJUInMO12BPUGTXGtOswIYFes7YrFO5o41kRVSnW0fHHy2m1yDORXt+Ec3KiTBQlyMG1agEx0NIYQKuyueRIShxJu0ewhaVC/6LKNUFjmdgdHcQyiUwtZ7Dgdp3qx1tM+vbL2sJGgjICJhRSENu5H7IxslhtYUesjPQa3s3+MMKQQ7LNlKiZbVKpyDy5N/E6sM+BgZsI4YuPK4PGlBdHOWGTTpxEsDlC8qjPEGmOqDq1npiOgL61f7tw4BCKJPObCBTpz8gSdPHeJ2hB6jFHJiWJTAASHMsZVV9hgDX7vcSvQmpfLytmdF4musvvp5DrhzMYCIuISlPjdZ7arXB1WSq1aE/OijPEFvsGqso6pk3M6PWXygcpam7a4zMH0LqkbQplpN6kFXugpLTI2MCTRPmyYaYCfaOstTJ2NGOWxqSkJPqm1dwE6PJdxFE0HAf4dAgetmseatrBS5n2FOpa3WOzWoa3zoMmDNrxRjFK50lpCmtwQBG4UAmMopAFq1oCp2Qohq5SjKoQfJegEGqtuZ3jwVQimvF2lQhO+WdrJS2vS8LEvtbBjK8XMjk1pWOVB6DGgRQvlJXCGgFFAu1QsUF35lLFwl0e9mmUoLZQJdXtQsQvtliYPDBrbm9LwbYuwyYJtGipB01IuoqxA2wUhqB1q+NRjbl/wLgTFSq4EXw0tsXm2FezO4jIKVIUQqWgFnU6HowVl8H0InXns/gL/VWl52rQmrytUsDDY1cSxUNERsnQxy/+7XL8wE7V6tHLAaklb++jk3ZZsBo6l4ZNnl9BHHdbMClC0u0qRfFYLjczYHZg+nQxW5F2ylm3nkjHCQhuskpCOkzB3o7NgquYy2MTT87HgOwG2br1A+ZoT9sby4wuaRxjQO+euuQoAMzDjhIIiGsLlQti3sJtXDUFJWxOBKBYs6QzcB3gpt5oUCcYVm+GCfBH2wWpSta4GGuSkOpXKGB/o/wQWMlaVBQ8ktwE/yK6QlsAAadYD/y4IKLVWr+Zdj8/uLAnqwh82mbKLhkpcZs3umjTBW+q1oEw8q7H2Pt7Nq34NobfvTVzC9SMOnlfpaoYTcOGwq+Bf8HkzVQb2bcUiJuCdHiwRVcxvNb3gEpGMNHDN2GBxCfeDztTHjtcEfF5Bvkpug3fPuqo/LPBibNpXif1zo8wkJAkCWwSBm7cIHZtIBo7YKCQolzWpwRMSDvvVHMQUCDpWj7YC0xXO9kU8S5JZwYccfkTsFN75GLK5EwJDpFoLNEQQ0iLwpXChwQm1GpEU5eETxf4jFWim+CgLA75DzKoj2TSVIHylyljhYcVcqaQUw4hlSvD7yOFYDl79RrGKhSCotHsGpbIx+GCY1IQ5NAkzHptmWcBS5jRsHe98AAxoieCrkcdqlJlOPA/tnmKeeZzan4GWD6Y3MF9e5dfKCYW9Oj7Dq5JdgV8bjgDgjQLJZFnVm4NJIcmFD6JNafLilDdUUXoTA5zelfDHQhY4ZQIrVGacTTjPF+Iws4Yv6yz4d0C/QFxhjWEmYP42pKJo2NZl6XTxAUJpRdQOn5l8HuZm3jgAkzLTE4VGoWWnsJkjYMR4J9KqUyx4zuQM7IP+McLtZtoMZIrA5I3+LVZt9FML9FYoAw0YugwO/0WK4egMs8YkwVRV0Rq7geMLPpM2jmsImqzNmaEzs6ozjqMFmFL4Q1baGKd67PAdSTsMAYPncZE+2wjbZeCIoCK1shA6yuBBHrTrlSpBKUw4i4IqrRy0XhWMM2hX03C8D7LFoSXK56B9SmIMepjV+SqFvvcEv9skL1qgySo1qos0fSa0aAn4opl1LOaghU9j84GltGkQjGAsTMdw5EUK+iis4hLFiqZjURcMoXdJV6EtaEgGc7Zm6ocR+AgbmQx5vLBSt+KUK6UpnYNAxv5n2AyQQZv0+hU8CAJXDn6lphulOI4CSsIPV6UojicqYz4Ck0qUbSAJzNWs4gkZLNT56JIUFtmuZ8KyoXmizij/CgI3FoGbeGvoRpIwPz9Pzz33HGGrKt15550bWZWUvdURcGswX7SogF2UYp1bh86ysWgo4IPW2X23DmXukCIeeughOnXqFN166607pEUjNIPNoB52+QbrD7tsUjZSx8KIRZFxSBCyYPmIYAGrBboGFkt1yiGiQnIcmi9t3HQE3nrrLbrrrrsIR3AQ7U0trf+1bxO9fYHP4Fj6bMQ742nuHBEceW19EXChBYuXiiKgrQusMJkrP7rweIR1KVQK2cYIeDis1czpDUisZeVjN4z4OJnubKqk053dnuyb2o7pI4q2cbcK6WOOwBiaO8e8x29g83l3VekG1r+zqoZpCEJaYBXaWU2T1qwJAXUuYT0LH8mSOpMwksBRQMpPYVwSTKHlJGXgG1GPsjYthmtsyhqX5ks7dyQCIqTtyG6VRgkCgsDYIYBjZ7LVFmXHruHdBkeT2LxjyVJwjIfAjmu6mDt3XJdKgwQBQUAQEAQEAUFgJyAgQtpO6EVpgyAgCAgCgoAgIAjsOARESNtxXSoNEgQEAUFAEBAEBIGdgIAIaTuhF6UNgoAgIAgIAoKAILDjEBAhbcQudRo4vR6HxyY4riZOr84uCtUyYiFj/VoLh90alG+NNQjSeEFgOAI4xb+Gg1XjmCtxHHwdx+HTZXWCviRBQBAYRwRESBul1xE2KJu3KYuwKhZO92/jPCKvmCVEL1kx2TWcRN/AEd9jlxBoPo+IBk7YcJzwjRh8RTnMe+xGgjR4VARw9h3ix5YiRUTZsKmNGLqtskG1DE7EX1c5rX9ujkrfer13o+tfr3ZIOYLAxiMgQtooGKsA1wgxEh7cHUtTLoVwR0H8uGFFOBBMRnhtWBHb9JmLQOAWjtQME0JwxWIIy7RNmyNkCwIbjkALMTmhbS4mg5BqHI4SYY4SiI1rr2fl/XNzPcsepawbXf8oNMo7gsDWQGB8hTS7TnlEA+ZgukYyh+DAQ7RdiFkXhaDWa3VIImZiVQXMQxQ7BCHPIRBwMpnE/2nKq8DGiKuYTyLep01NxJJLZqqdOHqdrodpo5rFM+QzETevgBPCdYF8H/dMBNlG/M1sra0DriNMegHmj1I5i3h5eM5041TtEseogyk2idPGteKqjRidSSogjp16j8uAeTZMDjSBKZSt6E1x2/WTJoL5pUuI8YdYfUmYdVOFZkfI4jx8z0Q9ZramgzirIMZ8eKa+Hzex4mcCPNCUgpmmbVEpjfh+HAAaOcqgo2wHVDD+wEy1MYl4lgh2rJKDEC4JxD8tcRsRjB7Bw3tIDzLLH0FgmyCg5rhJBlYnUcRjLTTswYS7NmYJzJyLggTEoMFvIX4vZ8Mcwpwt1EpwvUhSiacVYrlWmFco3oO5VurOWc9C/FgzjjnELhoBf1l2bmLaDeAJ1EBMzwyCmMMEm8C8X6LQc3musxsI5mo8oea65qSa1s58Jw7RlEOg+OV4A+pvsjtJwA/SJcRA1jB5dcTuzCIYei4DnoV6EJ/UCln1IB6CcGnJVIEqiJebSqJM0N9CIPYs80LQmUF8XebP1RTiEje73eHW0ojgwaBKEgS2EAIcu3Mj01NPPcVBq3zE7tzIalZZ9gX/yPSEoqvz/9S8f25hUDEL/rkjs/7k1AH/2LmrfS9d9o/PzfjzYebLx/25yYP+6aCsc/NT/v6Tyxd8+disP3XwtK+eXjrqz0wf8S+onzP+9Pw5fX/hnD8/PesfvcQvnfcPT+32Z/WFz/l3757zj6lnl/xjs1P+/Hn+rds3ibIVtZdP+Pv3zPrHL+P31ZP+gWnQFzTj6skD/uT+E+q9M4f2+Lvxm1/zUe+hqZluvR0arvon9k/5h85wucf82amD/pmgeRdA9+Shc5wb6ZJ/dGbWP6YK615r0hmzKf9gSMSFI2j7YbQOicucQDsCPC8fn/O5HcsjGJYtfwWBxQjs27fPv3Llyg2GBXxjfnIxn5mY9o/yJF8uqfk0r+fBcs/VnJrwJw+c8C+Fc+7ITIcf+AuY9zNT/uGABxyensZv/eLCuXl/KuAvS+bmEJ7gnz7o79k96x9Zwvc0gWcOTfqzmgGBt5z2D0wGfIZpnQ35Fr+LZ1OH/DMqWx9vWDjtH5ya0/yJn2LOh3xx4eR+f4L5qeJXV/2TByb9/Sf4YggPYV4KnHXbkWf/Hn9i5oivoLgKXhjwZ8V/NcPEgwU0dVrzNa5KkiAwAgIs1ygZ4s4HffrVLyz9//aPqefXk/7dFpIXN48Uu0H1V95dXN8/1zsapaWEIAQPtEqtSoLqCDli5utdE6bXooaNFWQYMTyG1azRpnbHF2tpaeEdq9WmOFZ3Sh9nQPNUSlEUKzyOuZfMmPp+BCs/E9q4Vrh8TCDsiaGKiEGDFU1kUJ8qACtZrEht9UhdZ/JcHlIsRRmYTJqsMWs3yIpDUxas1qMprB7bMMmqPGhnOqVNLRFo0uIu2aodUYpFHGrbTEOU0pU6FXhlH8tSHSvUMPJMHPSQO0LDoRFsttGukAjQk6ZW16QD03ImwDOG1XPUcYLVuSJSkiCwTRDAJoD664tpffcVqg3Spnk8vxQ3QGpCa643KiWgPVfKH5XilEG8VkO95oBXeJSCRkglRBxIYTJaSittQGvXhA+oLi+CzU5xaOqWaMI431CegGKTecqbAcPQNXX+NaHOq0MDr1IU2nzDJWfZSvoy9l6i/hb4XzqmbxrQwhscdzN4J4r2p1T1UWgMDfA45jHDeUgEWsuMansUmy+gnURMT30JrX/MIRs0xsD7Yq1GUA/aYTEfH0KnPBIEbgAC4xkWKmpQbDfQ/nkP4hOYyAGTGNQPsWSBGu001fIZmBba1ICKPgHxwXUaVEhCIFL80CO77VKa+e3Q5JEHRhHtOGnFyEwxATaYHEywPbREDYhubsD5IojJF/JxvB1hM0pYT899isR6yojCH0yRhv9Bb7MIxl7RnwPPIcs1O0JQNBqWhsLwn26GQdlyjjJZCG4wRVSKMEGoulA3hNRqqUYtC3Tb8EGDgLViQsNdmJm7/mkQAmPwU+EmMp0gttMm/Opt1oplywuCwJZBIIZdmhNEr/cuCHeT0Tu5e2lVExt8Qd2DORM+nSWIECWY/J2Qn2DexDqTA3MZghe7TDTVPVy3bYoU+GXMr6hDzUoFCykLiy1sQhgU6dUdzhMi4AmD5iDzH7tRoRJMsjYWU1abVh+WClKdx/6qIRaoL4rFXsDxwA663KDD/IbxEH4d/C/MxbCyW8uSBF6VIphg7QIE2DpZuC4NauiSzHJDENgcBMZUSEtTubyfWr/1Hfqpwnk3zZbKlGUBYbkERtj2wHBjmMFYrWYqDXKT2K0IP446ZJJIDD5VzRIEtjWkG8AUONB5swrC+1KTrwfQww7MDSsFATUL/7cmVZvw+3CrEPaqlAB2hawBv70CmZX+UuVaEBhXBOBPVi5RLfk5+mGwIJzcD4EmPWCS8QIMi6hQHhsJNWi8c9UGjrfpfxsCHnw9m5kylbAT3Yi2KGc2+l/qXA/iCQMzqAceNfMm5Z0clYtFtcitZzLLa+uGF3SDnsYpnXQp33Qo5bYoms73LA5vEElSrSDQh8B4mjsBgpGFyfLSGTp54iSduQBTYD4xSD4hpw7BpNDqgQ4CWzwCTRizU2asTuCwz6+41CyXqeGsNNa0RsxVZXACDbUWymHTYmhm1E9c6Oa7Gq6Vyg2egyZtquRraK5caL3UCpPpxso1LAaasEqhjtqHJDg/N3D8iKsE1CZV4nUqN3n1XaM2BL5SOgGtHVa/HS3AkLL4EWvReCXcIQImEqdXQ7BCfnksCGwTBCKJPDQ1F+jMyRN08twlatdhyhtEOzT8BmYim+JGS+Ah4Bi95kW7Dk03u0bAhFiHTqsMl4cENHfRCIS/QdLfWniCIhAmwkaM8pUcJeOwToC/eD2VgNN0m8H3B9WPjBFo4TrNBn9yF2nTl0FjnXhIPJMEf6+jHQRTcWyZiuSWIHBjERhbIY1hjxrwH+OdjIu3Uy3pkZgJ8x58r9TORU7Y0VSDfSHJfhpYySYNaJbCg4ycOnZfWmxtUIlV7VqY09e9/8axu9SqNzXvwk6lQqkORhWlBBwwmhCKFE+DEFVrxVDXgNX30mKDOzbVqy1dBsyxtXYc7cQj+KYkrCph06lKbgM7p+DDMrx0mFQKJeq4xaFRUV71M1fGWU4aFodq5YYyy3R5MZtjliMQmMXRrmbwsA0mCVMM3E0kCQI7D4Eo5l4aPlfwjRo+z0zKprErstjdoancEwYJN/ANwIbI7jzng3CxO7vNtbCQAxOnFvh4F3YFOysHzM018QTuJl5Y2Z2jiDzwyCqYhKY3BhcGmFuDXdtuo6bq76Ye3gD/MRP8L1zYOnjXZh+6oSNhnXgIfHrNdpFKLkyfxtAK5aEgcEMQGE9z52qhxtbxatHGMRtxKrDmDMzJLNYor7gI/LUqRWphG3yijBWhB+f7CsyAATeOwznVhS9Xol6AE3FuEePhreWFVgEruBIEuQi2r9f08xyOwchhdZqEkAfGZuSrFPrmjkw6C4+RCtT5BTBN0FTCbyU4YoNCuYmPQYIq8CfxYKQt4hiNGJ60BxUeSVGxABpSKSpDjPSMIravsw6xqPKmzBo0aexTl6Y4hFXmxRHgksygTcCsgfY0C1xDmGAGqkDoy+HokBLTEIOGrrI2c/EgmuW+ILDtEOB5isVaPkemgaUP5ifzmhg21uR7p09PuxJFzMtMHsIav+nBZFelqskv4KgebHDKgomUDfi4Ym6mIsGibcncHMwThkMYp1wJm35yWGDF4O4AOqGYCvznIpQqFqiKG2YF/rEQulKGHRTXzxvgCwaelIPXfgV804smqVJJKYF2kVy3iJj14iHYiAW8GtHUCkLhcCTkqSCwUQjcxFtDN6pwLnd+fp6ee+45wlZVuvPOOzeyKim7gwA7G+coCi1dbgBzF7AEgZ2MwEMPPUSnTp2iW2+9dSc3U9p23QjArw5CZi2D8+aS112YFDBmCLz11lt01113EY7gINoLzUV/eu3bRG9f4DM4+p+MfD3W5s6RUZIXBQFBQBAQBHYgAvCra2EDgdI+ShIEth4CIqRtvT4RigQBQUAQEAQ2AwGrTk0jTat2+d0M2qQOQQAIiE/ajhwG8J1rNXdky6RRgoAgIAisGwKJEjZvrVtpUpAgsO4IiCZt3SGVAgUBQUAQEAQEAUFAELh+BERIu34MpQRBQBAQBAQBQUAQEATWHQER0tYdUilQEBAEBAFBQBAQBASB60dAhLRRMES4o4RZ7juVv4EwUmkcwsqphbAsBuVbKxTmIBi5ydH4tmZycfhZBGGelj1/tofkJg5tyzS2ZhuEKkFg+yJQR5DxLI3F1GojjF40M7yto7yz2s7eiDJXS8Omvz/8+zQq3990sqVChYAIaesyEHAgbKtFxcS6FHbDColmcFp5NRMGS1gbHe0KZUsrSatrK1pyCQKCwDZFAJFTStlKd4GKA8IbFhatw5ozyjvD8vOztdS7Upnr8XxT+WTf96kPk3Xh++uBiZSxLAIipC0Ly2pv4rRvnLiPA8K3d4rgZHAOIn89yW5Ryx58Tvj1FC15BQFBYJsigJBVTcvuiSCASAqIhDCc24zyzgp4rKneFcpcj8ebyif7vk/9mKwH318PTKSMZREYXyENseLyCGYZRaBeI5lDbMvrESwQb89E+BU7wBixMgsphIIyU5TG6rGURviUULnkIZB6KYtwSSaeZ6k6yPaJiVTNIqQKgvOZKDtba4PBuQjHZHTLQnUeAjabBUtV7DTyCNGEEC3JJCVT3CZNj1WIUwohp5imdAVEejhdG7FQElx2AiFZWoGBs5FF4Plm0AiO94ey4wgvlcpQoYxQNT2mUA/xP/Mc9zSRoFRBxxq0ESIqWWiQXc+r+juxToMS5Y8gMHYIqHlskoEVXBQxKgsNzL9VJ5daZcxFxQtMSiHckyqllad4r3uCVaA4wsGVOjzFo3omod0weM4zP2HegP/TpTA+KHgXrgu1Eg50TVJJs5IuhS5C3CF6SaWQQfg65hkZqlghr3Sonkc4vATzkTiZWWji1aO+Mr9VpnQaYeDaNYSpAgZNvMTlJgoI0c5Jt0/xLZSVRuxjVcyidziKCuis5ME7mS+CJ3aY5wA6rJXqBc9sFtBujasJGsNwwl4d8ZoRtq+cQ7tN8PJ0mTrN7qKDAqANTACf5Xj6sjwc6KzEJ4dhDh4d7/BolFUGv684iiJ2Q0kjdms/X1b9EX6flsOkw/eHfV/QR6UM3HX0+EllKsvj0YuN/F4fBDgs1Eamp556iuMh+AgLtZHVrLLsC/6R6QlFV+f/qXn/3MKAYi4c9qdnjvqXFj0+7R/cs98/qe5d8o/OzPpH1QsL/umDk/7M0Qv67QtH/dndk/78eVxePubPTkz580FFl4/P+ZMHTyPH0nTp6Iw/PX9OP1v4/9l7F8CoymtffFEimZbYDJKaEThk1CgjPhjUQmyhGZTWsVoJSkt81dh6NKfXHqLtqem9PRr77/+a9tQaPZ4aPW0NreLYUhkpraOiDC3KgA8GizBolIEDONgAg8Y6wdDc39qPyU4yz2Qmmcf6Wszsvb/9PX7f3muvbz039i2bpba/68Havlm3a23jtnU3z+27eR1+HF7Vt3TW9X1PH1bbOrxqaV/Vosf7+HDL7TP6ymct63v6XfXaR48v6u93C+ZW+2Cfcunp63GeG0PZdW/f3Cq9vcPKnEovVttbd3NlXznaVu7B2G6eMVebu9rGDGVAUgSBsUNg3rx5fYcOHRq7ASg9f9S3cVnVQDpTOqtPJw1DB7eqb1El3rnBFzYuU+iP+tYf7lu1dEbfosfxZn/0dN/SqqV9T2sE5N1HavtKqbSv9hHtRe/Du8k0Add33D23r1YlULgP9G/ujL7bmSYptKu0r2rp4327YhGiw4+AZpWjP7XNw0wjZt2tjOXwKtARjcaAuPU9cnFl39JV3EiMNtHOxbNu71O65GI8ZvrK1/hWjdY9yN0NuEel2UwvFRL37uPAqraPpxp/HIPbGHQM/K6fcbHSBpddoMczNHr8EeZWaqB/q5ZWqZirVfv/m4Cmx6Phys3AMS6dTIC5cp9O5EGAewAAIABJREFUo3nM+E7UKmDxtyAeXTZ+n1Bx8FoY6H6878tH/AzqNF95rmfgG7dL6beYC/M1Cg8xZU4ffeGOof8qzlCuj6QUpyQt6CH31p6BXO5Od1TyFIv9jfhbsfPDjkr/B3sJ5BKPUYLk9yNhOKRnSrHVU73dINS3OpHQVz22YIdmDoViJBHGrgVbOgekXUpNUw3ugQTOF4HUz4kdNKRVSuN+2HVoKU0CHvIjwbHDrFwgMxIq2wO+qA2IvRGJ3C3qNTKjUihIyv4Lu1l3e90QO7SI30chB3aRSntmcjY4lSTsakEi6DrtGDt3py1MQXUzF60hPwQBQSBALvfugTD0bCVXmtK0IOxdTZBaqxQF7yKkY34PZFCgC3X2AHkUyVkE54JUc7GdArimSKICXvKD3tSYQqAnEXJCCq8UEyTrDhPolE7AbFQPSZnVQKbUilp1MxK816tvvxk0wRHyKVIUM6R4fpduw2qBNM1C4XBqbRrbj/gDFIZEXyGTmFNTWzNheDEKHJaanCqtsoCO2kETMffE44jRjH4KNNNnrYezhnrCCo2B1e+N0kwzpIY6/atxWEEy4xC5mDQ9Pg1PMKLoJVMczBPfO3K6HO/7EvR6yQIJpgoV91NDIZ/y4EnJMgLFmXHAbCVLOZA9YkC31Eo27WWNhbmJmRmoF6zRi+zd2R6jKislLQR6pRWzwhNFixm2a/0HcWwywhQKQw1rGI8Z9hsKAbRB3ByC6iCMsYS95AOxamaCFka/3hYQ4na1zUiI/OEahVjzZbNhECZnC7VBnO+w+6A+aKPGmhgTj0QUO7v+oZrIZCCc/e3hJP7P/UgRBAQBIwIWsllLiXYbN4TlZDW+2CkAFg6FoSrtpxomC36HeXNnhvrSTO3eIDZbYTAsNqgAnTBl8IDRcJIVWUfYW9uMmuFwEGpXJ3mVZnAcCJKpWXtrYfLBTcYtbG+rX4T9ktkcIZAHJirYJ7qozQVmEAxMEDZn1KxVTNamoTOmayYDfVIYBS4gdwOKYjulnwFd5d88DvQVdxyDmhhwCFwjxrlhDGbgqndrNoJiJH6D24xJ0xPQcMN2d3BT0eN4mMe9Qb0wYroc5/sSxPfFD/MXp1v9CERCYKyRTkuBP8mY5PLIEChSJq2O2toWke+Gp+iAgl851ba2gekaGZjRuzXapx5rBC1DTRNh9+wIUDukag7YvpnqmAirxeyETQWI9ODiH3wCrGZdu59qvGDQGuzkhk2Zp8E6pNaAEwPmhCvyZibGS64KAvgYN7S1kstxK63XNoRVi9hGNXMvjxV2YuEWH4XrIbkxO6kREqmAqYG8wQjZvBHYqzJRC+J9tVNjhwehggYvC64Ns4TZHrbVTC1tTdQC7olDOTQPs62R0JOMjmO448+V+0b8aMX/vtS0QANVP+IOcgWpvBlHcao7sTzWBjcFdq2jVY+vonU7IDZvso+EThgWnHel2EWF9FMhw+9Unwu0YRqoQgwHsZvWdps1cAAIuL3k9RDUFhpniZ1eBDvSKC8FN+v2Zs3AeFC3IZ8LRJzI4mgiN5wNwm0wAh1UxwRDZ97B6yXCO8zBjFqq05F6gkCRImCyN+Fd20HrVj1OqzbuwnsLlVqaWLBEp1+NCOERv5eQ3iifSzBlNUFI1CE1i8CoG/oAbOIgVYOK1BOEWtPGlSAFh3GD4XWGc0+rstFLqYCY9VMCpgOq5M0HCZq9uQ2mGDZIuLBVjKo6U2o1Wsmk0S71BByWPG5oAWK0Ae1Av8YxjO7gjIFuhz0OjNlknBvGHx4gFYsxhpRPJabhSZuJg7m66P3rpkg0M1xifV/4exDq/6hBlQ5HtI7BX40MD0SaUxAoWiaNJ2+2wsuJPRQNqoSRPxdWeB6FyOVS7ULCvo5heI6aQXvN5HVrtiVguFw+Czxr1F2MyQHPIx9UliGoPq3aiO3wsoLHpSuoHoc98EwCpYu174n42qjZrVXk6qzCGDxxEH+br13z0AzCbk33BhtccehxBNxcFmjH0I7kjCCQDwiYbeSog+1TjTXm+5hsClZ4H4bd+oYrTB7QFrvKfeHWGjBifmprD+KcXWnKDs/xYAcSh8MUokY5A3sx/HB3+NT3El6HLngBBlIcTSTsoXaPyjWFvS7YcTkU+zgzwvUEda6J64D4xGUawIiZcDEWXTDDzteEHafyyWcv1KY2is0/gg7pc4AWwQX1Lhz0E48jQb9wjQWD6yaPtqEOeVwUBB1V+NoRl8Q0XJlqAjoZD3Oy2Mjk96t2c1hHN74RaZdEmKCxWN8XK+ynQy6XZq8XIV87e8LG+rqkPRq5IQkCxanuTALKSC/XtHRQfSPCUIAwmu3Q4zsg5UqzUVsjXKkbG+Euj5cGN1ubOgge1moxO2DQ2kCt9jZSyTJOI3p3W5uXGurs1I5dTwRXWuDqbcEljQZFR2BtaIVTAgyAPbBWAe2tQ3gOveloJUQ+b2/xUyMMRFtNUI/CUcDqjV6N/6MGTG8T3MTB5LV6gAMPQIogIAikhkDYDZUlHIq02tZGF7kbW6ijrpEaICljYmCyQwoWfbFgxO00waDeQi6VI1M/sni/I20dUTbMDprUWN8EZg3SJ1AHM8wkOrT6yQZmAlNmhQTQ0cpMlhV2rAjVg5uskKLZscmtccPGF8xDHRyd/Io0LcbHm50cbC1Ub+MQG16EhDD0CnvfjoZGhETC/MC4mJugEma6ofKF/RWhsnWY2iEpbEY1zLsVvxmouONAI4n6NYFGYSxM49ox5Ajoans7sq6gSZDcEZeENDwJnYyHOeJvUCucN+rtHsSZg6QUz0QMtBOPfTAmg2vH+L6YHG3UofTrRtgqPAdWDjuSGXZ2cPdyPBCBcewamk1Qli1bRvfffz/BVZWmTJmSza5ypu0IPCcjcE4Ar4TCsWccFEAAotYoR5UzQ00wEFbZmkB8NRLAMZlaHeSHw0HaRCFBL3JJEMgGAvPnz6fVq1fTpEmTstF88bTJMbtgA9sK7/axJV8cJ62RzDDzaCz0jV/OYF48j/lwZ7p//36aOnUqIQQHUbVzaDPbf0fUtYNjcAy9luKZolZ3pohR2tV8LXAPd4XU+6Cq9AatZLem3czY3hByUT1s1nS1Q9Dnh+f+8NQ1YzsR6V0QEAQEAUFAEMhPBETdmYV1c7TAo6seNmLtsPWC/t/WgsjXuv4iC/1lpUmoOznadgOrXtgoGJ5jbR1ju5fOyjylUUFAEBAEBAFBIEcRECYtGwtjqUPsorpstDyKbZrIjvQz/qZR7FK6EgQEgdxCALavHDd37IuNmuHBWhQlZzAvCrRzfpKi7sz5JZIBCgKCgCAgCAgCgkAxIiBMWjGuusxZEBAEBAFBQBAQBHIeAWHScn6JZICCgCAgCAgCmUbgoYceolNPPZV6e3sz3bS0JwhkDAFh0lKB0ttIVqdrSNieVG7Ntzqc2oXz/YUTDDyVOgluj3kpG23G7CiXTgYQ1gDx7TxxxuSpR0DMtmCcq3JaEBAERoKAzWajBQsWUEmJmGaPBEe5N7sICJOWXXxzv/VAOzW0+qLjNNcjEjkS6SZyRk2lTtKJD6PfpG1moELQ1UDNnkyEskxhMLZGGGW3k0OvOggTB/KrJs2pmkI3UkUQEASGIlBbW0u/+MUvhl6QM4JADiEgTFoOLcaYDCXoI1/QwJQgRZRVD2Abb0Cp1Il3r35+OP0mazMD10PIeRgYJR4NcU0QNRz5A+NgYkJAZI5+IkUQEAQyj4DX66Ubbrgh8w1Li4JABhEoXiYNOduakPjNbDKT1dE4jPyaWAXkTutocCDVSg3+OajBFVDSiQRacdwaiC5TqB2pOyyN5I2eQdRsO1zbUTkSdFEjcu05HMjDiZycTW49CK4LaVQaqaO9nmrskLgMXnRPPTKEID9ngxPpp+xUU8e51LRKYS+1Io+f3V5DdpsdaVj0HKAD21yOtFGOZg+SLSOFlbNRzdPpQcqXBm2kyvzQPsZWgzRPzXoHxjocHRuRwNubOf0V+qupp3a/xuXEGUcwWb9QtvraMG8FV+QmRCiQoDY1L3Jj1SHvYBPnXEUMN2dz7JyiEWRGsDI+iPXmBD72ujbShxV73cLkbnJQI/IPepsxZ5YoDsY8wVz9zUiL49K5O84yYadmpQE/NSMfYHN7M1J2IXYefjfqa8ztISUORziIhYmvSWsz4qF6az259eZR39dopwblAUKuQ34GlecHuWhbY+MxeCpyLAgUOwLBYJDWrVtX7DDI/HMdAU4Llc3yr//6r5wPoQ9pobLZTZpt7+i7e1apMq7ovxnL+jZ+FKeZdTf3VV38eN/hQZd33Tu3b9ayjX3KbR9t7Fs2q7bv3l34vXFZ34yLH9Hqf9S3amk5+pnVd/cOrYEdd/fNXcTtvdv3yMVz+5bpHb/7SN/FVdf3Pc0NfvR438WllX0XP7hlSL9KK08v7Sst1frDCLbcPquv6uaNyqV1N1f11T64S/ndd/jpvqVVtX2PvIvfsdp8+vq+GTevU+vyf3Fcdb16/O6DtX0zrn9and+ue/vmzrq7T5mCoU7f4Uf6akvL+xY9zh1wd7hfqxd3HFob8fpl/GbNvVftC7NftXQG2lfRX3dzZV85sFN6A+Y3z5irYq703l8+WrWor5SxVG7jNqqibcRdN9TcuAx9rYrzICSY65bbeYz6fYexrrP6blfA2tJ3+4zyvrn3aovPaz/jdpzlYWG9Z2m/+XjQWihjUdo83Pf4ohl9/cu0o+/2uUv7eJg77p7bV6s8dCgf4bmeO6PvdqXx4i7z5s3rO3ToUHGDILMXBASBrCLAfI3CQ0yZ00dfuGPov4ozlOsjKcUpSQt6yL21ZyD/vNMNaVo6LDWkPZAsOZBUWFFXIWltfQ2kMJxHCVKnmqBXS6nkJ6/fRhfXhnBNFXWFEJTRBKmTGSmjPEFIupD0WCkWSEKsAQpowjQTguI2Q2IST+NlhvRLTbqOwLPIcGAK+JVk6jXIE+rWs7GbkZTYijycmpQtWZtGBPy+ANmQd1QZHSQ5ba3OmGMxmSEB1BI+m5GE1xHyKVKrROMw9jP4dxAqRxMkZTblgpmc9XbyRyNqIrEykr1b+BISLjttYQpqeA1uxwypnlMBzwxJk5VCSsUE6za4gRjH8eYao2r/KZON6uvU2RASTNshF4wz5DjNmMlRZyWfjgGeXx8yQDhMITyDEXI6rep96MfpMJHfr4tU4zQnpwUBQYBYkrZ+/XpBQhDIaQSK062FbX3KsS5HDGtTaiWb8uVPtTDjA1Wp4R4z7IvCYXwgwbA5bc3kBdPnNINhgQdfa6OX6j0+ikA15QdzV9NkRUc+Coc81OwA86NwQhEKBsJUBwZHKWZzTKZIu0oms6XfnokTukciirrVhMaCnnZqdfvAwITIj3E06DclaVOvxmOJYCpmdWA4baEaZxyALEh/FR0UfptxLwaScBz9HQ35FQZHabb1s6YmNswKh5S5cTFjDmoBaPi/fn5wQ0o6K70gPZdaEqybyvoNbmbgcZy5Jr4J69gPkFI13pjjtWPBM2Jq94K9A9POkdehvjWjlXA4qKikvUr7OA4EydScbuvxepXzgkDhItDR0UF33XXXiJJfFy46MrNcQaBImbQ6amtbRL4bnqIDykqUU21rGzXE4UHSXyxIPiDCafeFKGwBI+ZohiTHRNYWL4zSzeQJQHpmR6vMyFhgw+VFKIbBnQz7Oxshb1MNNYUaqa2lRWE83fX1CUNqDO56wLHO2ySsFOtihsdh7GLYY4o1zjw5Z4XNWQSMf7iRLJ4wbPGsGHhQkSY2dnhI4fmlCAKCQMoINDU1UV1dXcr1paIgMBYIFKe6E0hbG9wU2LWOVj2+itbtgJqyyd4vlUppJeCZZxqoagsHIQHSRCYWqDMj8B7yeYOQQIEFgyqTpWpeqPL8VqhDlT5MZIqEDKqvMHnb2siToi4sbJAuESQqYUVKBlWnx0JN7Y3ksKnegREWa6VdMDYwQ+Gwfi8wcvliq+kgrQtH28dvMKIW8/DHwRIwRSKplQjrao1Sw7TnYrwh8bolbTrmXPku5hyHg3PSHrUKYOwdIfJ4vOQOsqRW7dOEFdFV2Xwm6G7F5iCb40h1vFJPEMhtBJhW2+F8JEUQyGUEipZJ40UxQzpRx16CBtVa6ouFFxzSMq9b95z0kctngaelJuax1ZE90EFtPiu8C7lVtheCB2GrB8FidTs2thfzUodbY0hCbmqFJ2ZCHadhgBFvB7kUhi5CfpcXQhXYueF/FnMQKk71Qx3xtVMHPtpRXivGBCO4GOuzboP3q9/tVa/BG7a51R1TIhcJe6gd0h0uYa+LfMAVfrNJxxGvXys8OsNu3aMzTB6XH1grXEkGSuJ1Y8aUxxWvxJ4reHCILANe7VkAVq5hMkrxMOHx2MHs+9tayG9zapJXqKDxbLk7oEbnCvDGdcHzNZDmdiPeXOW8IFDICLRhQzxp0qRCnqLMrQAQKE515zAWLuJrRsiHVu1OGLPDPqi1EaEgGiGxckCOgq+ktalDM+TnamzU7qP6UBt1aHybHcxZoM1HTW0WrR0rNbS3kA92avY2xMuKwCi+HWEZUhTKWJiJaoQajJksSwO1t6uMTGNrHdU1IqyHBXZqYBbhU0CheHxHDZjUpgaywdmh1dNB9QZsOIRFM+btdLSCcTHBDMqlGfMPBNAEpszqhsNAKzN7VoSbQBgPMAoJx5GgX6ppoY66RmpAWBMG1gTc2zXHhIE9D+/IlmDdbM46CjcgZIe7mVzuxiHzjT1XjKOuBSFYGqHWRpw5SErt+GtOd3iDMLEOut+EEC22QANszsCZacXegmeuvgnMmkmxUTPXdVBH/+V0RyD1BYGiQYClaHfeeWfRzFcmmp8IjGPX0GwOfdmyZXT//fcTXFVpypQp2eyquNpGnDQrmIggYrCNaeFYX44Atfpj2NWN6cCy0HkxzTUL8I1mk/Pnz6fVq1eLpGQ0QZe+BIEiQ2D//v00depUQggOomrn0Nlv/x1R144ROacUtbpzKKJyRhAQBAQBQaAYEHDDpGLx4sXFMFWZYx4jIExaHi+eDF0QEAQEAUFgeAiwc9Lhw4eHd7PcJQiMEgJikzZKQGe8G6eLgjGkqxnvJ1mDZqS3gq9DUZRimmtRLKhMspgRaGhoIP4nRRDIZQREkpbLqyNjEwQEAUFAEMgKApxxYPny5VlpWxoVBDKFgDBpmUJS2hEEBAFBQBDIGwQ444BI0vJmuYp2oMKkFe3Sy8QFAUFAECheBFqQkWXXrl3FC4DMPC8QECYtL5ZJBikICAKCgCCQaQSsVmumm5T2BIGMIiBMWipwehvJCkP9/kRFqdwkdQQBQUAQEARyFQGWpI0bNy5XhyfjEgQUBMS7sxAehEA7Nbjt1GGIRF8I05I5CAKCgCCQLQQ4ubqeazlbfUi7gsBIERBJ2kgRzIX7gz7yBePlfcqFAcYfw6xZs6i1VU+3Fb+eXBEEBAFBIJMIcFqopqamTDYpbQkCGUegeJk0JMFuQu5Ls8lMVkcjubSE5GkhjITWHQ0O5E2swT8HcjcGlETXgVYctwaiTYWQuslkaSRv9IyXGu2ILzaEr4pQoKOBakA8amrsZKtpJHdIvcnfbKM6l35DmDqQbLsZXQRR39HsoaC7iRxOrX6ccVGoHflHG6m9tYHqkEfUXtNAHdFhhsnXVq/NpYacTXqCcyJvQw01tLVSvdOJ605qRmA0d1Md0kGhDWczeVkP7Gsim6OdtOHihJ+a7XXkSqIjXrhwIZ199tnKJDkCuN9fLEHXog+D/BAEBIExQIC9OxcsWDAGPUuXgkDqCBQpk4Zck3X1dN/6nXSk5wjtXv8QXVXXTL4hTFNiIINIJN5mbiWvD5IsbyuZW8EABQnMSg2FvT7Nhi2Ca37qOYA6OkMU8JEfSbiRE3tgCSFZNoRKrT4/+fDPXe+nphZfwkFYGzrI2+oka10beT3tVGcB4xZnXEpDAS8FwEy5PT7yNIaopdWjMJbka6FGVw118Fx8HmoMNVNTlMMKksdrpTaPh3wdDvJATeBCGx7MscPuoeaOIFENEqyHPCrDxu0FveS1OMmRJMv4PffcQ5deeqkyNHaHZ8IpRRAQBASBbCPAqs7y8vJsdyPtCwIjQqA4mbSgh9xbewYCt9MNaVo6WELyBBGSo76GFF7LVEP1NUEwbGB57GDAwKSoTJ+fvH4bXVwbwjVVrBTyeckERm4I/2JpILcP0i6NebNBokbhftlUaqNLMC5uAMxhvcYdWiCtM4dCCpMWBHNmqq8jm9KJmZz1dvIbUgnU4Br4P3CgkPCZMVencgRpnI3CAR6jHe0GyO1VWD4KeT1kguRNraWcSlo4uGRbW5tSb/bs2cSGvVIEAUFAEMgGAmyTxtJ7KYJALiNQnEya2UqWwRuoUivZ0uEoICcLhaEqNdxjtpqJ88Exw+a0gTkLYOlhL+Y311NLYw2EWD4wRBHyg7mrcVhjPBcmMkf81NHaRI1giuqgxlTZuhhV455KMC6+x2wxMIdmlcHE6XAoPMCI1mQBCwkGUWW5MC6zQexnwrF+qPzlWiYI08DYuXmOYfK6480x7sAH9L9o0SJyOBxKZVGDxsdMrggCgsDwEGDTivvuu294N8tdgsAoIVCkTFodJDaLqDIKcjnVtrZRQ1pMWqIVMsM+zAy1YYjCfqg9wWzUOKCShAQtACbME7ATTMqGlmAH7LyaKWBroOa2DjBr6UmihjY4umdMUHna/LAri0CVGqyhulhzTHFILEXTmTQ27hU1aIrASTVBQBBICQHe/InjQEpQSaUxRKA4mTQAbm1wU2DXOlr1+CpatwNqyiZ7VKqU2nqYyWIKU9CgjQwH+6VRFqgzI16oPL1BqmGOzOIgp9mn2K8p9mgxOgl7XRRwtsJezg4JnRkSL5a76YVFVv1HMW7XTiUeV7z7zOhPkQJqJQLJGkvdDPKzeLf2nzc7qN7qI3cHsIXTwAh4tAF98Y7XqAYVwpp8KaSGICAIJEaAN4J9fX2JK8lVQWCMEShaJo1xN1sdVAe1osMG1V7axQzTMzPUen6VdYL0yOWzkEP3BrCBSQl0UJvPSnCkRLFBShYmNwz1TTgRi/lRVIyBgOYhGSJXm4dC4MuYNbNAFxuAA4LCpsEz1TXIyyES1hm6JOOKM08rPFTD2FkGleth8rj8mJ8tTu14pyFBrDNRR4sHjgyx5xjvzkTnjbGMrr/+emJbEi6iBk2EmlwTBASBZAiwHawUQSCXEShqJi2dhYn4mhG+wk4cW8dud1CzD2xXYwc1hRD6AupMB9SU4SZ4Z1r1VqHStEFyZur34rSDOQt4Q2DkYutVTc4WajG3g6lzghFppqAThvyRsMqk1bVQQ7BJUZvWtYTJ7jDYltWA0fQ3wKa/gVyQ7CUeV5xZ17RQR52PGhBKxIE+2i2t1FYfe5xxWlBOWxz1ZI0gXIju/ZCo8jCusRTNqAbVJWzDaEpuEQQEgSJGgCVpJ598chEjIFPPBwTGQdybVXnvsmXL6P7776d9+/bRlClT8gETGeNIEAjCO7WByIVUWumzeOl1zOpZXcrG8Y44MK4wbelhWKi158+fT6tXr6ZJkyYV6hRlXiNEgKVoXpikcOgfKYLAcBDYv38/TZ06lWjKHKJq59Amtv+OqGvHiNTqIkkbCqucGQECIa+bIk5H1hk0HqJRDSreoCNYNLlVEChCBDi5ujBoRbjweTZlYdLybMFye7gcegPq3LRt2UY+K1aD6rZqrMaQGGsjx1RaEAQKGQH2GBd1ZyGvcGHMTRKsF8Y65sgszFQPR4r6MR4NqzBEDTrGiyDdCwI5jgBL0mpra3N8lDK8YkdAmLRifwIKcP6D1aBMjLmwNyj/ZucPKYKAIFDcCKgOX47iBkFmn/MICJOW80skAxwJAsaYaqwCZSZNUsGMBFG5VxAoDARY4r5+/Xq68847C2NCMouCREBs0gpyWWVSsRBgoqwzaOwNKkFxY6Ek5wSB4kCAg2Tfe++9xTFZmWXeIiBMWt4unQw8XQQGq0H1eGsSFDddJKW+IJD/CPAmzZhlJf9nJDMoRARE3VmIqypzSoqAUYrGsdWYgRM1aFLYpIIgUDAIMIPGsdLERrVglrQgJyKStIJcVplUOggwc6YzaBzGQ2InpYOe1BUE8hMB3pzNnj07Pwcvoy4aBESSVjRLLRONh4BRDcoqUKM3KF/T1aLx7pfzgoAgkH8I8GZM3u38W7diG7EwacW24jLfhAjEUoMKIU8ImVwUBPISAd6M6RuyvJyADLooEBB1Z1Ess0xyOAiIGnQ4qMk9gkB+ICAZB/JjnYp9lCJJK/YnQOYfFwFRg8aFRi4IAnmPAEvRJONA3i9jwU9AmLQcWeJIJEKbNr9MtV+YnyMjkmEYERisBuVrHHdNiiAgCOQnApJxIL11++u2N6jyxM/QiSeemN6NUntECIi6c0TwZebmDz/8kB7p+A3tCu7OTIPSSlYRYDWozqCx8bF4g2YVbmlcEMgKAvwO33DDDVlpuxAb5e/Ur/Cd2rdvfyFOL2fnJEzaGC/NkSPv0y8f+TW997e/jfFIpPtUETCqQTnGkh5niYm+SNdSRVHqCQJjiwDHSFu3bt3YDiLPeu/p6aHlv3lMBAqjuG6i7hxFsAd3dejQIeWBf//9DwZfkuM8QWBwblAetjBqebJ4MsyiRkCk4MNb/o8//pgeW+Giry25gk4//bThNSJ3pYyASNJShiqzFQ+8954iQRMGLbO4jmVrRm9Q+QCM5UpI34JAcgRYksYJ1qWkj8CxY8fI9duVtO2N7enfLHekhYAwaWnBlZnKrNNnG7S///3vmWlQWskJBFgNqqtC2SjZqAaVlFM5sUQyCEEgigCH4JAYiMN/IPr6+uj3T7rp1de2DL8RuTMpAqLuTApRZiu8+eZbtBIPNouMpRQuAkZngpaWFmWinHJKiiAgCOQGAmyqIO/3NAphAAAgAElEQVTkyNdizR+fpo/+/hHNm/e5kTcmLQxBQJi0IZBk7wSLhp9c9RTxDkRK8SDANmqczJkLM2/8WyRrxbP+MtPcRICl3pJcPTNr8/w6Lx0GXbvs0kto3LhxmWlUWlEQEHXnKD0ILBJm0bAwaKMEeI51o6tBeeeuq1iYeRNmLccWSoZTNAhwgvVJkyYVzXyzPdHXtviVb9w//vGPbHdVVO2LJG0UlvuljT56bu0Lo9CTdJHrCBjVK/yRYKmaqFxyfdVkfIWIAEvRli1bVohTG7M5vbF9Bx2FKQ97fpaUCHuRiYUQFDOBYoI2nn3uedro25SghlwqVgRYiqarQdk+ZuvWrRK3qVgfBpn3qCMgGQeyA/lbb3UqITquqv8aTZgwITudFFGrou7M0mKzWvMPa/4kDFqW8C2UZo3eoIsWLVKm5ff7RQ1aKAss88hZBHiTtHjx4pwdXz4PLLh7jxIDlNMdShkZAiJJGxl+Me/u7e0l91N/IBb9ShEEUkHAqPJkb1CO4SRq0FSQi13nU5/6lBgwx4ZGzmoIsBT78OHDgkeWENi//10l1NTVVy2l8vJPZ6mXwm9WJGkZXuOjR48qol5h0DIMbBE1Z8wNymrQBQsWFNHsMzPVZ555JhqzLjMtSiuFhgB7Wkt2kOyuKqc7/FXHcuLsOlKGh4AwacPDLeZdLNplES+LeqUIAiNBQNSgI0FP7hUEkiPA0urly5cnryg1RoQAZ9Xh7DqcZUdK+ggIk5Y+ZjHv+PDDDxXRLot4pQgCmUKAVZ56flBWg+qBcTPVvrQjCBQrApxxwBh0ulhxGI15c3Yd/j5yth0p6SEgTFp6eMWsfeTI+8pOgUW7UgSBbCEgatBsISvtFiMCvOHZtWtXMU59TObc09OjaJp2BXePSf/52qkwaSNcOda1s85dDFBHCKTcnhICogZNCSapJAikhIDVak2pnlTKDAKcDpFttjk9opTUEBAmLTWcYtZi0S1L0FjnLkUQGE0ERA06mmhLX4WIAEvSJIXR6K/ssWPHyPXblZKYPUXoJQRHikANrsYi28ddv5VE6YOBkeNRR0CC4o465NJhASDAGx1dMl0A08mrKXAcUU7MzirQz11Qk1djH+3BCpM2DMRZVPvblU8S7wikCAK5gIAxN6iuwuGguBJvLRdWR8aQiwhwWihJsD62K8PpEru7P6QvffGisR1IDvcu6s40F2fbG9sVUa0waGkCJ9VHBQFOdaN7g3JuUP33qHQunQgCeYQAS6AlBuHYLxinTeTsPCxdkzIUAZGkDcUk7plXX9uiiGilCAL5gACHGDDmBl2/fj1t2bIlH4YuYxQEsoLAn//8Z/rHP/6htL1t2zbFXMUY0PaUU06h6dOnZ6VvaTQ+Aq9t8SuqzysWL6JPfEJkR0akhEmL/9wMuPLSRh+xaFaKIJBPCBjVoPpvUYPm0wrKWDOJwB133EG8WTEWozRt3bp1wqRlEvA02uIsPUfBNH9tyRVUUiKsiQ6dsKxJHiIWwT773PPCoCXBSS7nNgKsBtUD4UoQz9xeKxld9hBIFLyWpWi1tbXZ61xaTorAW291KiE6JDF7P1TCpCV4bFgs/vsn3cQ6cymCQKEgwLZq7FDAhRm32bNnF8rUZB6CQEIElixZQmVlZTHrXHfddRKSIyYyo3uS0ypydgLO4iOFSJi0OE9Bb2+v4iAgidLjACSn8xoBCYqb18sngx8mAsygXXnllUPu5nhpX//614eclxNjgwBn7+EYpJzNp9iLKH5jPAFHjx5VYqCNdqL0AwcO0PJfPxpjRHJKEMgeAiefUq08dysee5Q2bPgL/fzBh7LXmbQsCIwxAhUVJw4ZAas6/7LhJeWflNgIvP/+6DJMnMWHs/lcf901dMIJJ8QeVBGcFSZt0CKzLvw3jz0+JonSI5GeUWcMi+AZlymmiMDn5n2Bzj3vs8oz6F33PAV27KDGb92S4t1STRDIDwQ+VXY8TZw4cYA67bTTZwjtTbp8HCJjXNJamazA2XxYovb1666myhOHMteZ7CtX2xJ1p2FlWAfOuvD9+9/N1fWScQkCWUXA9MlPKu1brSeT7YwzlN+hd98Fw7Y9q/1K44LAaCEwfvx4OvvsWdHuBh+P1jikn9QQ+Pvf/658lzkNYzEWYdK0VedE6cyxsy5ciiBQ7AhYTz6FHAvUKOD+La+Re9Xvix0SmX8BIWCffW50NqeddjpNjONMUEBTzuupcAy15b95jDgdY7EVYdKw4gfee09h0FgHLkUQEAQGIuD88qXUdOt3lZO+jS9S+88fEIgEgbxGwHLSSWSxWJQ5GBm2vJ5UgQ+eAw9zeA7O+lNMpeiZNBahsiiVRapSBAFBIDYCuhrUYjlpgBqUpWxSBIF8ROCss8+hCRMmENujSckPBDgdI4fF4uw/xVKK2nGARafsxckcuhRBQBBIjgCrQfkfl0BgOyRrL4kkIjlsUiMHETjzzLOpq6uL2CZNSn4hwOkZWQX6uQtq8mvgwxht0UrS3nzzLUV0KgzaMJ4auUUQAAJss6arQVmi1vaznwougkDeIDAJYR0WLvxS3oxXBjoQAU7TyNmACr0UJZPGOm0OVMuiUymCgCAwfAR0NSgHx7XZVG/QMGw7RQ06fEzlztFDoOz440evM+kp4whwNqA/rPkTcfrGQi1Fp+6UROmF+ijLvMYSAaMa1O9/DXHWXhA16FguiPQtCBQJAq9t8SuqzysWL6JPfKLw5E5FxaSxaFTycBbJmyvTHDMEWA1qt6shDliixgxb022qd2isQR0Jh+mTiM82obQ01mU5JwgIAoJAQgQ4feNR2JYvuaJOcQYppFJ4bGeM1WFRKItEhUGLAY6cEgSygIB50iSlVVaDWq1W5Xc8Nehzz3rosUd/Tb3iwJOFlZAmBYHiQOCttzqVWGqcNaiQSsEzaf/4xz8Ul10WiUoRBASB0UWA1aB1VyxROmU16OCguJzNYNu2v9Lu3UF64okVwqiN7vJIb4JAQSHA2YI4pBZnDyqUUtDqzt7eXvrtyieJOexRKRNn0/W3XUJWdNYbWEWtv9tBY+eaMJ0uvfVaOr8Mgwm/Rr/4Tw/t00Aon3M9NV08VT3q3Ukrf/Z7eqNHPZz8uW/SLRdV4qCLNjz8MD1/YJjIjT+DrvneYqrGExbe1EH3PTuKKT3KT6HP115A9lMryWwqoUj3QQq9vZnWPvNXOjB2C5IakMBtKXCzAbfuVx+le/60J7X78qBWLDWoHlCUh//Wm28qjNpVV11Ln5CwCKOzolWX0Xe+fg4xmeh6/iH6r5cOqv3OuJL+z9dmEH8g9v7xPvrla/joRev2UuC3P6EndqY4xKlO+s43zkUfvdT5+3vose3JX8KJ515D3720Sukg+NQ9tPx1jUCNP4UW3VZPdpPad9eLv6D/euE99WD86XTl95bQWRg009+fgv5qd6U4UEO1Uy+j268+h0wUoW0rfka/fzv9JoZ3RylVnjOfHHNPp2nmMvTfTeGu/eT/83P04tt5wHgM9xkZHlgx7+KsQRycnhOzl5d/OmadfDpZsJK0o0ePKiE2Ro1By7lVP0B7Q73qqMyTqWKiPsDxZKliJkwrJZU0raL/WoVlsnoQeY/2dvVXy5tfE0Gov1FPC2dVUUWZiUpKSqjMXEnV532FbvzGBaTNLsPTAWGdcTpNFZOqpLga1aCVlSci1tqOAfcwo/bk739HfZCASyleBD48cIDC2vQrphneWrzLFo1B48vmykqKvnZM5zSxQ9e+94bPoI0Z7OOp8nPX0o2L5pDNYqYybDBLTGaqmDaTFl79TbpyRnYIzPjy6XRmVfmYzTobHXP2oF91LCdO95jvpSCZNBZ1sm46uLtwpBDpP2g9tPeATuYmGxixyWS1GAWoZrJU6hwcCMLkKJWjUPINb/rDyvIdlXMddBaLBbBrD216lNp+ch+53uhWei2xfJZqNAFipoZRWuWg67+9jBq/dgFNM3w8MtV+obbDatBx42KTH1Z/rvnDU4U69fyd1+411PZ/f0w/+r/3pC5FG+5sQ3tI32OWTa4knUKVVk6n6J5SeaenkprciWh8xRTtWoRC+zSJ4HD7H4v7Sk8nR22lIr2k7p3kfvgean14HYWUsZSRbQ5yjGZ0XJNp1hXfpO/+67V02azCYtIYpvff/0CRqHHax3wusalkHs/oyJH3lYVh3XSxl/Du9yCs51IG5kt7vSeCqJkHImOxVpISc3t8PzMXPrCPjuQdgOVkrdJJ+D565cU9dKTnQ+oM7APLNgiHDM2tbOqpZDUbmd4MNVzgzbAt2mApmnHKr776Cj37zNMFjkL+TY9jS45KfMljByioS/IrKqOMGdOqAW9bGSRrGn9hnmzWruFelbPJL4CngZZok+veuZG2HuihngNvU6eGQ0nFVBpEukc4v8lUPaMSKtXCLZzukW3U8jkxe0F9XVi0yRI05qBztowvpxm1C6jm1ClkqTBRb/gg7d39Om14fgvtgwFFv73YPlrzs+X0KpshlMJO6TbVTonCm6n9P9eSYipWBVuPr7OtB+zHfg77sUGbx2Oh3diFzVRs5BSVAexKxus7z97dFAhWkq3aRCbsVM30Dh2EukDlNyCFCmqNTYR910WfJdtUEENIqMJdUKOCgKx9aQ+pFhJQL35ftQMJvbiKXjHPoZrJ+8j9K90Crn8lSmdeSU1XzlDsPAK//y96YjtbjEDEf84CckA9aYGqtQzXuvZupw3PrqM3DrIor5RmXfcdquNJdD5Hrt3TUddE/t8+RpuGbJYnUoVOxWCH1qWZcBzbt4XWPr9fIeDhA5rnT5J1GNBv+HXyvFFK52OM5pII7d26ltzPvkmmC2+ixs/rTOFUcv7r/6azYL/jNtWrdn2R7bRmdTfZvnA69W5aTk+8jgElxbMfr0L+5V33fNLpvfTiBvoEpG0Lv3Rx0rpSYRQQiGNvVD5jIdV96RxIkiHBAm3YVrKAnGfi09/1ErU/6FVplTI8qO8s59CiuRdAnQcLta63ybsaeRhjGooegTQMEnAmOmDE2Fxj94flkPorYnIKB3ZSxDYDUrRKaAbG0ybsKCumamrRrn0UUozRBtIWUwS2qaCJr7yg0xaiqA3u4Hd1iPnXeKr68s3UcB4IDGjnmp8/Rq/yLnbidNBHzIfpoxn0vPsAdb66jjzJ6ON/r43aCCsT4qYmlkcZpq6QvkU+QoEX11GEpx05AAs1vXLqdDkSWEfeYzPxzQF97T1A255fQ2veOJHqNZthbtE061q6c8br1PEfz5E5Lr1lPC+M0upYeOrDy6W/HEONTZ++tuQKOv3003JpaCmNpWCYNBZp/vo3K3I8UTpe9IuvoXp+0fG6MSNkmjaVbNghWct7qf3xv9KRfftgi8E7JhAk/v4zsbBU0TR9pWCTMQ0E6wDOl8MeQyFZ3XuocwjDwvceoBA0nlZ0VzZ5MlrcQyVVU1RC0PUO+WD4a6uGcW7FdLKUbqRwVF0AxhG7OKIT6fNXL6GFrB6NdCm7U4v1VLJPA0NFv6SHdCNj7VGrgN3XZbCjIBCqoeVEOv/zzKChhMDkKQwaWKEZX6FrF81UjIrDe8FUlmGu1XNoCRPxn7tpp9Hy1zqfloCpLAHZj/3gsg2a1jPcsKOO2EfeoU0vvWMYUgrrYJyA+RSymV6mYLiKzp+GkBJzL6OFb99Ha/e9SYEQ1BCKkUw37Q28Q9sOQmanq1RNp9PCK0qIIQkq7aWHp3EIhfb7yiVfI7YbCSNG2pEj2l/8Vo/D1N3drUQR37Dhz2T6pInmza8tNAhybz6m8YpEnbdGKfttTDyDLrt8DlnxCkT2vonn/ByaNyO+bMZ63rnUuRcvdQlsriwzcC+O//vlmFL70N4D1HteGd51VcL/aoQZMoYNG6XXt1BkGpi0shKVOdt5jCo0g7TIgf3YtjJtqQNtmaHQlu4Q6GoZ6KwNbUyLQVuGvKuDlqf8HDAn6g4wvHUd+RUeit9n2L8q9BHMWfBD0K6pdNZF9ZD8pUsfAQkTCqX0Um80jEQP7Xt94yCGLj06YrLOJMuru6mrF5txYGC/dCEFO/9CgTd2UwVvPLnH8Nu0befbChMYldYNorelMxanjqc2k1z5w9JfzjLEAW/POnNmrgwrpXHE/taldGvuVGIvTk6UzqLNnC7jT6Xzz1Rfge5X3fQLeO6VaZ6WpupzyVb+V9oEW4y9kTnwStRUlLs/pMlTp6rMmDI5MGnYOb76NoiSZuTfu/cd2htz4iz276UaFo+BEawYvx2bUq1/MIN7D0wEQ8jSIdipVYynEBg55YEAwdnLTN/Uc+h8xX4Nkq/VkASBqTvzqmVglErIct5sqnxJk+hpfZeYeiFlWkev7ARBGGSiP3HGfJqnENhu8r/wGqk8ZSnZzj1dnVvXX+jRRzbSwckX0M3fWgDiezqdf2op7dxumFgJZHDB12jD1jepE8xnrBJl0nBRc5sYWi2VdTDqervfpLV/AqGEca1NkVxC+lhaQke2e2nt5FM1Ju0IbXt2jbKjnxy1e4PkoPtt2vDi69SJdaSpF6SF59CBF86Z4xBw8kRsMvhfrMJENRw+TBzolhk3ZtrKylQpSqz6cm7kCFR8/kb6wefTa2c8GCVm0Pi95uf/+X1TsBFsAM2J3c7e51302GslNPeG/0XOaahjUqVHxtdNv7NnHzsPnAqGx0QVldBp9mIzqRAo0LW9MGEAbTsftKgCUqyJ43vgDane2RU8AEZzIp0151SNtmymjv/20sFKB/2vmz4HhyKdthh3gIPeVaPhG6T5VZ+/QFVFRiD9W79f9dofQB8fBX3spRlXLKN6SBGT08dY+PR/jnsT2QOnSZd73/4LrXnhTTBfp9AtLPkHHTUdt582rX6ZrGcy/Wfh4EZ66lm24S4lXhalDKC3E8n25XTwjDW/sT3Hmz4Ox3WSxUKTJ58wtoNJo/eCsEljD76r6r9Gn/rUp9KY+hhULSsH86X22xU6qLzoRw4c1CQ+kJwxkTm2L2pPoTJheGmqWIzfTZ2dLKEy4ZgrloNwqY3tfXtfnFAfxyi0W5NqmeD5xFI4hcqxOhOELGqcy84DcBqYqlE5RU2K17Wc1Y9cjsDTkwlaD3XpzgiYC/i6AaU3uI5cePG37hzkWQUmsOYLqhStd+9G8r6tE0eWFqqEqbfroOrNBfVvWOGuQHyNXl18KrKT1vzWQ5teh2o2EREbOKyhR6msg/Eu7GqVIUWwwx3aWoIzYEhXP0HPv7aDduMrlC6eCRou+EvjIcqZPLmCTjm1ms4973xh0HJ0xU0TJ2pS7SPUpeji9L+xBtyrBRqNUDfepaQlzBtWtRbTQtYcKBQK6sy9Hxocoyog2WctgEJKwhQ6oOghQU912rJfpS0w1ejSaItZN2SLDmLguzpgbAYpWhdUmVs1VWj/+9xNXQp9PEZdIU0hyZ6mg5wx49JHvbN+Hm1A94MP0qUjvb0R5fvQ221kSge3GuN4AL1NF88Y7Y3xqXHjxtGVyEiQTwwaQ5biYzHG6KbQfeWJJ9I3b/h6btukmSZGbQ5Y+qcU/FV/QVWn8Fwf0t59ENZbKxQV5cTxJRDPY5ki79ArL0WoohqEqnIKlfLOUdntQczOUpo45Qhc2buhfytjlcGpp2AXyRUPUpDVmZpxrg0Srgrr6ehSfRy6sIPl13miiVWLaokON6JxRxisMl5D171HjhgPoyMynTqfzleoa1iRovXvmvvVkyW2JfSDf4/eovzoF/9r5yNHKJyEzijj1AYd9+FOaR0GjmV4R/hghfvvLEkFz6iOdng9yl2CwHARiBcnLVF7vEFWC+iYznelwH8lajN6DfRpL3aLdiu2pmDQbAgXxCVyYI+iziR2jIJkyASbNSuk7gqJgc2V6nAAaZE2tN5jveomFtJZvZiiqkX9zMB3NVoRxGQapGgKvwe7tbUv9nsKMn1Su6iged/63zSv/yb80vo30Kt49DF6WxRAkLBBG2Bj0ynRkfifhAGjTHgwgN6mi2fClkf9Im/6xCZt1GEf2uEJJ5xA32i4nlY8/gRxQLucK739b2yUgTByEhpx69q9XyU+7NVUUa6EdujtfBsi/mOKKvQsNv6vOKLuHLuxq4xlj6ZPnqViveciqCwkcOdpqsXoPTDOPYCdH+y/TFPPJpvCwBnd1/upbZQWw25FLbg2hBgPOaFU7Q0foQjSAzHLpxL1GGKwru3keXWgs0EvpIwDioGIDbygH0HipQ8BDJEmtITQ8WxadPk5ChHvfsNDv9+d2jrE7mMkZ9PFcyR9yb2CQL4jgA0rS+7ZqLbibDof7zQXRQug/GCziplQz02ms2aVqgwTAr+qTgP95ElnpVQXdhWTpKTEAF036FcZAsuWoAGFDg7ZKHZTcNNGChg2ZIr2YciGKzZ91LvqjUoXOT5a/w646sJryKGYUBygDSvWQssxSnRkEEh6ryPF0wDtqPwsRU7gpV9bQidb1eDIo9JpBjspCHWnEQ+OMHxDw3U0ZcpJGYQpQ03hodffW/UlRLsskVKah6G79hYcgwH9Xv5dBhfpM1W369DOPdSjq0Jh/FkN71AWpEXi2qNpY+55L+rKbq6oUPrqZQN97XLUi9NcoXl2QsqmBUiLRNV746O7UpPOrfFcEtMcrQfe3L5MGzp55mVk/9IFVBndJWLOGiC94Z3k3/wybXr1Hepio3/8S0klEu2Ff3zYL7kCdmY9qBDjhRfUin9wqFWkl6msw4CmM3CQKTwzMBRpQhDICAJRjYCyAeMmsTmK7o5G3kUXzDUUMlNSoXlud0HToImJNMcoFp2bNbfuCEIHqVu7ftrCXhAKySkBI6dtilNX/cGh6VV4rLIW0zSDnBedEg2e2/8+w3xkK2gX6Nc2eKQy7VL+xdiLJkLkww+xmdUqVETVsfBYn6rSLuu0coV+jw0dyRSeiRDI/DU2gWJ+IF8ZNEak4Jg0npQJVIJTQlirpmd+1VNskSNFW6CunKz/Ky+l8QgLEdbeQjNsLJhwlFdO1iQ+kGrpO7Ge/RpjVUn2WVPxYh6ggKLS1FShvHM8j89jMxnXHk0fKNtoDNzSsaRO3wz2gKgpqoNodbiva/rIY0cOai7fPBc2sChFvDXNeJttx1ImQjAq/vPrql1IxRy67Dw9cCKrGFROr4TtShgQy2y6bNFXqA7/aqYaxYzGQcb7fQQBjPXZTKXz506HunginTrrFM22DuE9eGee6jrE6ybm+eRjzRyeMQcgJwWBUUcgAhMHlbpUwmvuRJo4+RSqHmB0P7IhKXazxiaQCSUaP43NNlQjM62GamurHhhpixZnLRpvrRd0R7MdS2V4Pez8o9r2ls1aSPO0XWb/+1xGlqm8Iyyl6i+otKvuopn9XpKp9MF19kJbom18y2bMoVmguaWTZ9JZ07QGuoAFaG5W6EhUVRJvsBnEM14XGT7/6U8fr5hAsSlUPpfkX5Y8nd0EeI9dA/foUc3dacTKuoBuhJditITW0QP/vZleeSMMBgvu57MuoxsnHqQSBGfkEul8jQJRYy0OgaGqIRWHtjCcCTTNX9fbqirUDPF7Mns0tW84D8CVneBqrRaEithn8KVi9QAIg55OpReu6lGmbd/r9EroXMXFvPrSa+ma88bTNKsii8POEXkwtRZT+XNsHxwGguci3hlsPGoX0plvIF8ojH8Dr71NkWo4FZjPpSXfmKy4ySu2Jd2IHfcGM6bppUI5sMlL22Yhbhtc86d9/lr6rtFbret18gWZyiFWUkrrkHxm3fA87IXHbQn+zfvalVTyrJsC8W7LIJ7xupDzgsBoInAM3tb+8DmKB/m0i26k79bCG1cVmmdmGEewaQQpRNQMtbD5RnRz2AONJwgj7HTVEtZCB/ER05Y3QVtmkgkbw/obKqkb4X1U2vImvRJ1XkplmMfo4Na/0LZa0BVTBdV8+Vx65RGEDTG8z9aLr6dvntkL216VPu598WXanUrTxjo9b8Jz9ABVI3dySdlMqvuWMVREhDpf3K7a82aMjrDnNAYAptqEWHfXfxleoH8yhioaMDjgidh01WdmAM90gUm//qRJkyR3Z/qwjf4dbP9UD130ubPto995zB6P0e5nnqCVm3YqoTHM1qlkAgPWufU5evTJvw4wug/Bk1PXJkYQj0vfTR4Lva2qQrl92JbpzFvM7rSTHyqx17TSi3uMojPNOFe/PDDn3Xv0ImLLrEU8nVAE0fxhHBcBkdz2/Ep6dFCMtET9q9c+7JemGdQGPTvXUMdTYFA5UaiZU7wgflzgJXL96mlKi47qA/jwTfr9r1zKmLu62UYNqs1uxHjb+jT94lcIIKkQ+NTXIdm8eravpZUcg6i7F3GfEONtYqJ9TybxTDYyuS4IjAICx/bTM7/qIPfz2IBueo5W/voJ2mCkLyMeAptf9NtVdIGWGW3iw3CMiuoJBuUb7tn5tEZbwtCWwtShBExc4DVyr/hD+rQFDNSGV9WJlUybT86ZLDkzvM/dCLvB8de6dpP/j4/So5sTGQrHA+UYHXjpUfrFU5upE2qVCGhXby8SrIPmb/jtL8m1XZ95pujIe7Rh9XOI9Yi+wKlNQ07nRNQro3jGgyAD59nUiSVohZBcneEYh9ghfRnAJW4Ty5Yto/vvv5/24eWaMmVK3HrZvvDsc8/TRt+mbHcj7QsCgoAgUDwITJxCZyKSfQmH53n7Tdr34UQ677pldJkVECjaA8Q+LB40CnymzCqMy+k5sokTh+NiTdpolP3799NUxDGlKXMg0XUO7XL77xAyZocSmHu4JRHjPNw2c/K+L33xIsRamkjPrX0hJ8cngxIEBAFBIO8QMFWRY9ECyGEQyPqZX9KjSDdUbVFnETnIwWilCAKjg8Bpp1UrYTb6w8KMTr/Z7qVomDQG8nMX1BC74675oyRuzvaDJe0LAoJAESBw8DVas2kmXTsXgbIvvpma9TSr8Nb2PP9OnCDbRYCLTHFUEfpudcoAACAASURBVGCTpku/7KRPfKLwfCGLiknjp+a8c2crjJr7qT8Qp56RIggIAoKAIDBcBHpo97O/pJ++Op2sSHBeVgIbUHiFB4P76UMhr8MFVe5LA4ELauYSa8oKtRQdk8YLyQlWJyKdCef7/Pjjjwt1bbMyr2MwZh2f1F07K11Lo4KAIJCjCPQc3EM78U+KIDCaCHxx4YWKhqyQS+HJBlNcLQ5ux7HUWKomJXUEXnh+rUggU4dLagoCgoAgIAhkAYHLLr2k4Bk0hq0oJWn68zJ16hQlGvGvf7OC/v73v2fhMSqsJlk9/PLLm+ifpk8n2xnGGD6FNc/hzOZgVxc94/mT4sXzCY5wDtsIzhen/1b/8jHOf2I8nYAgx5w4XIogIAgIAoJA6ggwXeVg56wRK4ZS1EwaL7CemP3RFU/Q4cOHi2HNhz3Ht97cSUePHqVtf31dmLRBKE5Gyq3PzZtPrhWPKilhkpWFX/xSsipyXRAQBAQBQcCAwHHHHaeE2MjnNE/pLmjRqjuNQHFidg5+d+JnPpMufkVV37/lNWW+bzKz1jMky3BRYRFrslbryfSNG28ifp4SFd4JnnuuSNESYSTXBAFBQBAwIsCmSWyiVEwMGs9fmDTtKWBHgpxNzJ4D7+qH3d301ltvKiNhadr27W/kwKhybwgnnlhJ3/znRgRuRoDDOOW0006nT+F5kyIICAKCgCCQHAE9UTqbKBVbESbNsOJ6YnYOiidlIAI7drwxwGFAl6oJTkMRUBj+b9xIzIzFKh988AF1/OoXsS7JOUFAEBAEBAEDApyHsxASpQ93UYveJm0wcJxOgvN9PrnqKXpj+47Bl7N6XG4uJ/s552S1j+E2vupJpLcwlN27g3Ra9aljmupruHMZrfsWLHDQPff8lJ7+0x+jXZaXl9M/33ST4lQwDzZsK1f+jl7csIHubbtvtIYl/QgCgoAgkDYCe5Ha8e234yVgT7u5lG5gE6SvX3e1EjKrWIswaTFWnqMWX3lFnRKe47Ut/hg1snPKjA+4o3Z+dhofQavvvPMOvb5164AW2Itxz+5ddPVVS0fQcuHfetGFDmppaaG77rpLmWxjYyPdece/RycePnyQSuDxyevu9/tp+fLldO+99xY+MDJDQUAQyCsEfJs2jyqTxonSr7vmKmINVzEXUXfGWf1x48bRVy77MnE042Ivv/71r2MmiHW5XMUOTUrzZybtl7/8pZJTjpk0Y6mrq6O2tjblVDAYpEceeYTC4bDyjxk2/itFEBAEBIFiQoATpbOTQLEzaLzmwqQlefI53QRHNS7WwhKz3/zmNzGnvxXSNZb+SEmOwDe+8Q3EmHuZrFZr3MrMsDFTZjabFVwbGhqi+ArOcWGTC4KAIFBACJw58wy65up6YtMjKcKkpfQMcNoJVn+ydK3YyqZNm4jVnfFKR0dHvEtyfhACdrs9ZUwcDgft2rWL+C8zaLNnzya3253y/VJREBAEBIF8Q4ATpfO3lrUOUlQEBIkUnwSObjwBgfR+u/LJokqL1ItcnXfeeWcUJY/HQ8xsWCwW5VwFgrhKyQ4CutSN8V61apXCsHFhbydek6ampux0LK0KAoKAIDDKCBR6ovThwilMWhrInX76aYoYtpgSs8+bNw9eiPMUlFgVx/ZTzc3NxKo5KaOHgBHvZcuWKYwyl1tvvZVmzZqlqEalCAKCgCCQjwgUQ6L04a6LMGlpIqcnZl/h+m3R5ftkWykxZE/zgclCdXZE0AurRKuqqpRDXR0qDHQWQJcmBQFBIOMIsAnRpV920nnnzs5424XSoDgODGMlOeoxB9f79KePH8bd+X2L1+sVRi2HlpAZM13tyfaBRk9RcTbIoYWSoQgCgsAABDhWJMckFQYt8YMhTFpifOJe5fyM32i4XrEPKpbCUrQFCxaQOAvk5oozw8ZMNBdeI3Y2kCIICAKCQK4hwInS2XSITYikJEZAmLTE+CS8Wl7+6aJKzM7qzi1btoj9U8KnIjcuso0arxUXlq4Vo2dybqyEjEIQEASMCBRrovThPgXCpA0XOe0+PTE7B98rhsIG68ysScltBNgzVHcuYK9QPYsBq0BZGsqBc6UIAoKAIDCaCLCJ0A0N11ExJkofLs7CpA0XOcN9HBWZRbfFkJidJTK63VMGoJMmRgEBZtZ0uzVmsDlAMRdWX7N3qDBso7AI0oUgUOQIsGkQmwhVnnhikSOR3vSFSUsPr7i1OfgeG0FytORCLhyfS4/XVcjzLNS5sYSN7db4LzNpnIZKZ9I4DZUwbIW68jIvQWDsEOBE6exsxyZCUtJDQJi09PBKWFtPzM5Rkwu1cPiHdCLnFyoOhTAvnVHTmW62Y9OdQpiBy+dwKxya5NixY4WwTDIHQSCvEeBE6aziZNMgKekjIExa+pglvENPzH7hgtqE9fL1Isfg0j0I83UOMu7YCDBjo8dgY8YtnwPkfv3rX6f3338/9kTlrCAgCIwKAmwCJInSRwa1BLMdGX5x754/7/P0qU99itb88em4dfLtAktWWB2WzxKWfMN8NMerp6HiPtnuUHcQ0Zk1Cb0ymqshfQkC+Y0Am/5csXgRsYZJyvARECZt+NglvZOD9LG78ZOrnooaaye9KYcr8EdbAqTm8AJlcGhGu0Ojeptjsa1fvz7qLZrBLqUpQUAQKBAE2OTnsksvkdA/GVhPYXEzAGKiJjgxOzsUcHTlQigsYRFJWiGsZOpzYM9Q3TuUJal6/DVm2O+7777UG5KagoAgUPAIsKnPVy77sjBoGVppYdIyBGSiZvTE7CxVy+eih2wQtVc+r+LIxs7Mmm6TyH915o1bZQmbFEFAECheBFh6xqY+UjKHgDBpmcMyYUucmJ09XNhOLV8LqzsPHz484MOcr3ORcY8cAWbQ9JhrzLCxilScSkaOq7QgCOQbAuwwd+UVdZKHMwsLJ0xaFkCN1yQH8SvWxOzxMJHzhYEA262tW7dOYdRYJcpEWySuhbG2MgtBIBECeqJ0Nu2RknkEhEnLPKYJW8znxOys7uSo0ZJxIOESF+VFlrLqzgbsJWoMesxpqNjhQIogIAgUFgJswiOJ0rO7psKkZRffmK3ridk5yF8+Ff4Qcw7IfI6flU945/NYOd6aHtKjvLw8OhVm8EUlms8rK2MXBFQE2HSHTXjYlEdK9hAQJi172CZsmaMvc5C/fEvMznZIkmA94dLKxUEIsBSNgyBzYRWoLlVj71AJ6SKPiyCQfwhwonQ23ZE8nNlfO2HSso9x3B4mTJiQV4nZWd0pqqu4yykXUkCAmTJdXW4M7cG3SmiXFACUKoLAGCOgJ0pn0x0p2UdAmLTsY5ywh3xKzM4SNFZdiSQt4ZLKxRQRYKma7lzADNvJJ5+c4p1STRAQBMYCATbRkUTpo4u8ZBwYXbxj9sZpM5ZcuZj+uu2NmNdz6aQYgOfSauT3WIxpqNjOUXc84Gfs1ltvVYLmyoYgv9dYRl84CJxksSghNo477rjCmVQezEQkaTm0SGefdWYOjSb2UNggnEMsSBEEMokAh/DQ7daYeautrVUYNH7WFi9eLKrQTIItbQkCw0Cgqmq6MGjDwG2ktwiTNlIEi+h+thm66667JJxCEa35WEyVGTZdDcpM2q5du6JMGqehEtu1sVgV6VMQEATGAgFRd44F6nnaJ0s29AjzeToFGXaeIcAqUN0DlP+y7VpVVZUideNjlrqJSjTPFlWGKwgIAikjIJK0lKGSioyAhEyQ52CsEGAJG6cl09Wis2fPlsDKY7UY0q8gIAiMCgLCpI0KzIXRCauZ5MNYGGuZr7MwSs04DZWe4J09QyXIcr6uqoxbEBAE4iEg6s54yMj5IQjwB3LVqlVRL7whFeSEIDCKCOjeoNwlp6HSvUXZueWdd94ZxZFIV4KAICAIZAcBYdKyg2vBtqqrmgp2gjKxvERgsBSN4w9y4RRU7Hww+HpeTlIGLQgIAkWHgKg7i27Jhz9hVneyWklipQ0fQ7kz+wiwJG369OlKR+wlysdc+Pldv3599gcgPQgCgoAgkCEEhEnLEJDF0AyrOzl+lTEIaTHMW+aYvwgwk6bH9eN0VEYVqYTyyN91lZELAsWCgDBpxbLSGZonf/TYy06KIJBvCLCTAWcx4MLSYM5BKN7K+baKMl5BoLgQECatuNZ7RLM9duwYXX311dEP3Ygak5sFgVFGgCXB+gaDJWr33nuvcsySNvZaZvs1KYKAICAI5BICwqTl0mrk+Fi6u7uVDxlHgJciCOQzAsyw6eE7+DcHyNXV+DfccIMwbPm8uDJ2QaCAEBDvzgJazGxPpby8nPbv35/tbqR9QWBUEWAmzegMw/HXFi1apIyBzzPzJir+UV0S6UwQEAQ0BIqQSfsfalz8C3rokOEZmFBGM6qnkcM5j1oX/ROZU3w8ws//hib9sJNowlm06rmvUh3u89x9N13iiRDZ6+jwfbNTbivFLtOr9tbTZL3RR+abbqHWPb9Qx2WpoY1PXEI13NIbuP4tH+0mEy39yffJNTe95jNfO8baDOjETMt+fiu1DTcP/bsbqabeQ5uojG7++b9Re9J2PibX/2mlqzb0DhhF+QkVZJ/noPZbziZbKbwGYzwHqWHzMYXejZD5pOOxAkSRd/9KTa1ecneGKTx9IQUevICsqTUktTKIgO5owE1y6A7+x04H+vlUHGfcd/1/tPiFgc+NcYizbmki/1cnZXDUudDURxR8l8h60ifTGIzhnrTfzzS6iVYd3Xc6tOlpanggQL5QN1mW3EiBm08azqDlniJGoLjVnWDOqixlVH60m3ZuD9BDP3uU6p46PKLHwVQG1YnFTDMs/Nkd2xLY3AkGzEyOOZ8Z24EURO8mqsS6VpYRHTnURetXryTnA/9DYMeHVUJb/kL13/wpWX+4jUJaC96ONfSQv4sOUAU551nGlsEf1qwK7yZmzJhB46IzbIU3y5HO6CPyPbWKahb/lJyrwyk2FuOeCSayTFfpp3VCis2MqFq23+n/odaf+eiZPWGKWKzknDn234QRwSU3jwkCRShJ68e5/MIlFPz+yUQf7KLGb3XQQ3sitH7lK+Rf9EUarv+i49v/QsFvj8laDur0MHk2dBGdcD7VnQYpTS4MKekY/onaV91F7Vwvuqsmqr3tu+RddHzSu7NaoXoeeX85n2x0mNr+5QG6dXsv7X7hFfJ955+G9awEXvDRE51YlZn6qD+mcLcqeSmd4yD3NXgupYw5AsY0VDqzxoNiezaOuaZ7iw4eaN2d/059d/LZd6n5unb68R6i0nn1FP7/z1CkpoVVYKu62k+boJ2YkfLEYtwzeTa5fzM75RZGXDHr73Qv3ml1lDVXL6G2+elIGEc8O2mgQBAoaiYtuobHn0xN8yrooRVgavbsJf8H0FYyT9BzmNwPe6h1814KQNxhqbZQjdNBbQlUogPUnT8xUcNlLnrqaAktuqOZ3Bcdp6jGLFCR9uhqxyR9eO/5D1qwuhsEfiG1lwWgtpxCV094je7091Ll5TdSCExCP0Mzje51/TM1sUT9YCfUZkTlF9oU1aY31Qe252/kengttSlz7iXzdFYDY85fVdXAgcf+k854GDhNt9O9c8LU5glR5AQrNd22kOybPdTkCVIQkqD6m5ZQxyWqBC/8xkZq6vCTd3sXhcugKqyupqZbHFR30nGpjcq4qwZD3fYzD8bXRSHgarFUUM3lCzG+k8nCrR3cRa0PrKX2zSH1Ovd102XUNDsWgez/gFZd3kCB75ycwgd0EjnsUIhjLgQJbKgnzhQSjDP00H8qa6qU7R46+YuddPPMTkjR1FM9L7ho3KHL6N37PkumLRp2ncBugplsM6up+ZaFGnZ/o9ZvPkDfxzrXXl1H9u0byDPdSW6LJ801wnP+kIdaXghS4FAvmU6YRnVXO6l90Ukp4BFn/gV42miXxpk3dLUn260tX75cSZmWsOA5Vhm02OsW+M5pFHnrZWr82Qa8u90UgWTJivevrsFJLXMn4d7++6ouxDNw6BXq2N5Npmo7td/xFTwT6D3R83/wZXJcsYbW4/28+RYbBbAh9XWb8P7Ukfvmk1XJbZL3n/7yOJl+EKCeE/D+X91LrtUf0D9N2E0r8Qxy2bmincZt52fXDlrnocYOVvVFyFRWRtZqm/Ls1ld1Kxsdfm4H3NPcS3VDzBE+Iv9Ta6kZdMUPMwC88LAPPJ9ab5lNdpgaRMcDw4Blt5SRd0UnBY6WkbOhjjo0mqX2kui/qb3TidYmMOSd/ivNoX20+aja7/rWVrIduoUC13yGwone6egaWejm26optHobma6+lhyrH6B/AX2oxHcAinJq82vrfpuN/PhGtUECHzmhmlrvWEKNp4GuJnoOEkEh13IOgeJWdxqWg8XsasGHV7NX83c8SotXBmhTN5iOy61EnZ20HCrRhuc/Sm0hS6dBxM1VeymAjyyXAD7u/F0vx8eWpXWp9tGzwUs3ePbSzkOfoLPnqGM9sAeMCNqIgLEMcOMWMFSayUPYD+J41ESOedY0PrQfk+9hF12lzNlMzgungWnFnB/gOYNzNRZmZsGo1SGw+4E9Afp+06PUccJZ1FBtop5DIdzjJQ9P9OBfqeF7HloOpsoE4lpv6ab1GzbQ4u95QWrSLR+Rq9VFt74Qgs3W+dR201lkObSXnnjARY1/4TX5G7V971H6/gt7wWWDsby8gsLbt9GtP3CTa9DwuefIJh91QMJBE6ZR09WpMGh819/I68fHggvU5Rb+UAwpicdpmTcPTJkmT7FU0+03fZa+umQhLVWD5FMpPrr3LrGS+eAWqv8BsAMBNs08i5zAbtMGX0zs1q90032oF9I+CsqQUlwj33920OIVAQpMAJEHptZDQeU5b9ry8ZCZyQkVAQ7hoXuH8pnDh1UziYMHD9KDDz6YMkwD1q1nFzV9bw0tB+NlvdBJbU5+fgP0Yzy/HbD1MpbdnSGyOc8i+4ReOrD9FWpcwR7XqT7/XeSCvVw9njFLN7IwrHBTy1vcehrv/6Ft1PzANtq0p5esX66h2hPU0ZXjHVee3be8VPfDV2j9nhKqu8lJzaCDWzf76Kof8HtfRo4lMe6JgVpkk4ecP3uFnoHk2nbhWWTr3kvPrHbj3FuDtAMh8h+qpkZ7GfV0d9FT2Ki5DsZoMOapFN7pJGsT653+1k12mqVtMKvAXLVgbJTyOx2iDjDrT7F9qmHMBzr3kmmeHeseod1+H13StIEIx46yXjoCOtz08DbUT/U5iAmGnMwxBESSpi2IaUIJ8fe2BwyV+qF7l1ysLkSZCylFO4x8vUeDigTE80InhS86O4WlPJ4cMyvAiXVRcA92OtirBng3SCVUM28amKd0+iijRbfVUcsc7CQh4aok2C4pzNlnyQIG8AhaLbdXQx3H5WPyglHpAfNRbx8krQr56IJaX5yxhwxzXkKur5aR5+hP6ZIXIphzcOCcp9uoadFsskzw033bg1CrgkB+dTbVWALUshk77aMRhbiE/X7ysNCo7Cxqu+MScoLhDfBuHgTF9RbUylDFplOsMNi/F9ythRm+07BD3/AKbfJHEJQUvUGy58VHg/G1zrFTI5jMyNFtwMhEJk1wFe3r6GHqWLEN9l+8O3VQg8bcxh3Lng3kvA5s5aEw7dbUklUX2hUpZSxVcqJxWr49m+qr19JD23GzgpsNcoAq+ttqLz3B459+FjXNx44bkgivjt2PFpPz0EYKQNKwNRZ2E6x094+cVI/nLbJam0VKa/QBBZVnElIe9NvwVQuYtBJlzWx4F4hSlHbGBa7wL7BUTc9p+/rrryvJ3i+99NJoaqqECBjWjfBOOyDBtOF5rXHiXcImwLs6SE8cDZEPm4lGwzNaNaeGGi+pINMLG2j9Zrxn/Ez2dCV+/qMSadAfbAoaLwJdWg0mEDZTvu1gMk/rTvr+mw2TsTohrboam82qCLX9yUfrsbm1zDxfeXZp9zRqvMWJjYyF6hadTNaZIWrDu3ogFKLAB8dR/UXnk3PFoHve1URr0T5AxzwB5R0tnbOQ3N8HzX0aBvitnXRgwzby0WnkiNadRg1Xz6YGbFxbX+iCLS7oD787kxOgn9Y7jQ1vorWZO/idno13uozWQYOwFRsn2zzQqzM/iXd6W+J3WmN2lVFX19Cq286nGmx83do7zUxw0yVWsm720jNgtEsZ70vOwPptoCfwberpxryTPQdjbDmSYEXkUgwEhEmLgsIfJK0o0ohuCrKYigsIi2cTiIsmYeuBpw5/1owES6s55I9tjpUqoUZlqVegx6wQW5AySNhY/bY3aR/RBvEBbQWxU5iwMisI+Cv0FAzY/e9+RBZWvTHjN4cZP5SeILnBuJTOtJFjyAvJxrJgWrgeGKndh4wsRkSTIrIaEbs+fKDVvxFIx9Q5823GYkK/euE2mdk1ljCwYoEaYecX2MzbdWZWuYQpaPSwHXBXvINPYo4WSPs2UOsKLzVg7ErbSsH6QXJZZzfRU5thW/izdqrZDm9dqIqadLWqURrR6YOKgNfcQk0NpyVfS8YKNouMcyVUUDVz5lHbt1Xp21AmLck49SEn+RvSsTtBk9jhLxtUb2U16yDsKi90UPNc9SuuSFUNJfEaYSMBJ4VSf4iObFhJ9mV2ar4JnqtnDnlwkoxWLjMCCxYsoPfeey9lMIzrRnQSOe17qflhPJtwIgmB8VKf794hGwH9PVPXVqNdyZ7/qGQJ73cZM98w1OfXHCUMlSRvN9TnKv7730/zIH2+CYykwgQNfQOoqpqcnVBTrvRRywPddOSoNkb8jVFbHcSQ/2LDrNEnMwaq9I2/TGd6+B0wSsdZlcxMqNE0Ykh7g06k806Xpr42ibpN+k5HmbQSqGyh0j6T16l/ouq6H6fSb+5IU6EPoLvJnoNEA5RrOYeAMGnakkRAPFSCWKK+7D2RKDHZtNJFl6w0rl0ahAZqTYWh4h1kCJI0/rhOt6pqyXT6gE2HwlhxOR5qzWqip8Cc+SGhM3cyAWTVqmZ3BRUJS0Ls86pVOy39Pv5rsZN7SAgOrYJhPMZblN9pEdf+u3VjeDoUgNpxIPsQOcrqtHQkNR9Qxw8fpVvBXJWylOhqK2w2oA5UGF8ux1PDj24k08Nualq5lzZ5PLQAxv1Lb2P7OKN9VS95Vvqx24b0EarHxir9/gR/qxfSDsVxIJWSbJyptIFPnyax49rq2uuvKwySFez6i6ks+nSk1rihlu2rDeQr8yBUAHb9/lfoX761jdrZTuk7Z0gIkLTRTO+GAev2wQ5qbFpDT+DdZUlZ6zzYfD0ASbFRhZ2w+VSff8NzY2wvnfcfmzGzxuDFGlLkjbXk/CGH9ymj2ssdVD+hEwxbUJH4p15AZ4fMHfQZDTDTFx5yLfWWlZrpvNMjXht1bKm/08B3Qjq00Tj39J+DNJGT6qOIgNikaWDrKh+IqcjKu5lSTdoEycmiO35Afevv6v+X8sea29Hs0o5CpbBBtR2rBOOmfOyH3QcMXWfyvhJqvs3bCEKQAfZovg1QB8A42DEnzThMPB5tJxoljvoOGER5OGyAqUxjLKbPoy1GDNf/O7nnp0mEWEII2xRmWepuWUKtXz2LHEb1gILpZ6j+2/9MQde1dPc8M5Ue7aInWldqNjdcgQvUobBBZPV2BBJChi+jJZVxptCh+YT+PZQifYgyykMJ+HDWpn8InyT7JYvJ77qFVoHxrcRztRUhRup/N7JwNClMseirDFi3zoCqCqNqaoFpQJOzOn0mOaXnv3+TOYAJyuD7HwAN4k0QwYOy4zvzYXoA1Wzaqw2aozGCvIlWiv4OJGES0+4q2Q2ZWBv0kc47nWxICa+n9BwkbEEu5ggCwqTxQsATr32DapdDrMpSRPhQBSgMQC9UknztY+IAleNq7yTTsi0DjDkTr+XxVMN2aVDzeWAndkBRS1o0gjX8PmxQbZaj1cALYNLwt98eDUawmzFeGKTXpSIhGjB4MKgKUYSaQZtzEEwMl9ITzEOlcgPujX3A6lLFth62XEEWVe7+C9mB4bjae6npjdj3xD2LnbOqKmFpJzN4unpGu6NnB9UvvpvMl6Lt0GnUfMdlVKfMBxLHzoHOHrYlTmqwQG2yfS01/O5vcbsc1oVk40yxUTPUmwOwA4aqvSTUn4OZ0xTbHFoNUr9/+w9gdjfZV8C77uZ6ap3DzCGcXRQ1upRRQ0B/bnhDxJslpGEb4AiSbCApP//8frMKLRw14TCfwCzUyN//eJs73gxpFHbILIZKy/Qqn8RzrrJ2RzQTkxDeAUUah3iUsZ12hjSfmRMjXRttFKPyTqf8HGQGGmkluwgUtbqTbXCsS0sozC++IjqHcX5DjRb3ykL1kMTctzJMW1dAqrDHTH4l8ryJnJdXK/YR8YjO4CWzg6Fiu7Sdiocnghra9XAQyfsY3JZ+bGLv0AnbYKzLNl8ldLFujwbjWzfUf5WXq6E34t0f+/xJVIdQJD9e2UVb4S3YECqD9E+b84XWYeyEsXO0Q9pV1knPdMPz6AeryN3dSVu5czZSh8o2rXI8wneAsXpmD2IsrdyI8ANB6tDEYMpOu9QMJjMCL6cIwoj8gczVMKRmHhMOFP2Yaz2WnUwtN1WTC+FQNj28htrn3TDAMDutcQ2unGycqK/a+mFwna9Q82MV1HrNtMGtADv22lKxa/7hH8gNlbGKnQ1OE0OqD/MEwjwA0yOw49v6wlq492PNNWkl26pJGUUEsEG0wRFnExyU2ju2gJFCzEZm1phBGKTejjmqNJ7/9SvWUBO8SF2KbaMZG0lV6j68998E1afK2O/egGdoJkKGzLRgEwm62umn1qdBK1fvjQ5ZZcqG3tMWjRnYPzsHQghVevx0wO+l+nv2UmRzULlYOecsNWtKf9Xs/kq6Nrr9rvGdHjqkpO901G5w6L0pn0njOUi5Tak4ZggUtyQNO9XdkBgdAftRhXhay+5oINdFusH0cVRz07X0OBiyuSd0k/uFIIUtVrr+tgbq0/NzdgAAIABJREFUiNZJcd3Qdo1u0Dq9mhxRj6MR9AG7NKcWsqHfEQE+DvCs9GM+DniPDqfU3FSvzrksTC7EJorAfm7pLfWYc6w4Yyn0gACVHT9CeAk7pIkIh+GCS/4s2IE9/RN4dipionTKSdR82zyqheNDCLZTHYfs1Ho5s8vs3cY2hCdBTVRHN6Mv0x4//Rgx3Dh0xd0/qVdjxw3qynIRQgNU4+OCj2JT68sUTGcoCesmGye8vS530u1zKhQvXTc87GL2PfnsKHYReMm6QiaaC1ulx3/kGFYA3dhDPo4cWPN7ndOoqhvMATxeg2XT8Jxfm/5zHrsDOZsqAich3tlNNpqBcAp+2FL64dHYrGxkOCiqpu5L2Faqz7+JLsZmLMCxteBBOnfJZdSspUgb3vsPGyjEP1tUDcnvoU7qgFe8CSEn2i60UOUEPN+wRau5yaF5Ympel2w/OuieWA4FprlO8txmp4uBgw/MWgDy/Iudl+FcqiFzEgKW+sUU1iZ33ulUn4PUpy81xw6BcX0o2ex+2bJldP/999O+fftoypQp2eyqyNr+mDx3qeExCIFxt8AZwE6Iz/VvP6Wr/NW06smrqE4c9IrsmZDp6gjMnz+fVq9eTZMmpWmXWcgQRgOlmuh65OrtGPNcvYUMtsytGBDYv38/TZ06lWjKHNhfOodOefvvYG2zg0bCZhW1unMoovl0JgzPTnXv2W+PhhxxJ9hoaQPUZMKg5dNiylgFAUFAEBAEBIEhCAiTNgSSPDlxMEieTh6rIT4a1H0N3/8qNeTJFGSYgoAgIAgIAoKAIBAfAWHS4mOT01ci2zsRcZuLHhg3p4crgxMEBIGxRqAMOXaRBaAOGzt71J51rAcl/QsCgkAiBIRJS4RODl8zzb+KIutzeIAyNEFAEMgtBBA7q+6rSNkkRRAQBPIGgeL27sybZZKBCgKCgCAgCAgCgkCxIVD0TFokHMvxu9geA5mvICAIZBOBSBhOPdnsQNoeBgIIByL0fxi4yS2jiUBxM2n+ZqpvDSh4e/8fe+8PIzeW5A1Gr7GVntjAAEqojaIxQHHOKVpbNLq3aBwgekpjPyjPEj9LaXXlZymt3fSUwBnKa0d5lvIs5WEM5bajxIcDxMUaoieOM+Ja4hjb4gKDFte5Yu8ZvF88kvmfWVmlyqqSKt5MtZLk+xPvx/eC8SLivXANcie7YqM4c6yBwNUcGuArTum4SUbL+6we+i0c1Dpe9x5C6pkI/F0cXrvQSNjDIbcu7ez1fVaPFgsnwKj+Dc53iy6x0l1WlY6poeGQYBysaxZ/do+9IfE+rIr3sUt6vsS6U4/ardGZ4ceqx/4X1OlkiIOjOyoKysYUjajlWGTZvcK3dmPuHT2cUKs52vpQ8h0RIdUKAhsRuMVCGgQnCGhOywRAIflhnSzz/NHlNqI7fRiQFxpkG9vlvq5c0ahFnc+QdEIvIMNmPC+aIvKDiveQBGDmYOrrDsHXWzQJ+uQUr+9z+3FR6s8uF9Kgn5B5hBBZ64TNsyu4+hw46T008SHFYbpB8ed1LJwejPeRQHBb9z6unsob3WI07FPabJ4Rh3PD2L/RvVsiDuMlwmJqM6sD7233iLoe+V4HszqmSbuNw5qvoaNJfKbwfA1USZOCwBSB2yukBQMa4LCKpg4s+IOTmusFgMsYLBE+cHWcEr8rGfAyaASr8oYBafpFiYzxIa9B0M0jAFyIpJQ//Ahnxe9kObGwYFjrmX8NcfzqZbuf24/lhi/vOh51aOL0qG3AzBKv0xZeXluXVVMMwYwsazVua8jv4zIjH1wWxTesHmjReiOdXOeMebFp7N+wLm0iJ/QRE8AyN4eQS32ahCY1rILX4HqIxSGHJ7vylMaiSbty0KXB8yBwS4W0lCb9EUxzjorBSSwAINSQ33GpAXWXYXcgtBUwJj71mza+U4jpqUMb1vGKSR3TAFojt9+Dytwh0zCpOYrKQuR1HAgs0PxYCFjdn1ACU1E96uPevCkAYXgcm/rh/CtLKRi0qOE4ZJvz7eH+sEWODVpQl+WOpqGEAmg2Gp2+ooPbbI2hueuhLzAnGBa0TGetUGOYHlBv24tp1LLJ5X5U9huPvB5w4r4ZZDoD6CGR+CMTgy5eQqfQGKFtVzXskWs0qc+0Wwi7lRBF4zZ+5+V1PBuWsOE9BPweWk0816mOZyWkEYQFTQup02xAI1kno1G0i7fBpuQ2ZAla1495aOEVdOXYlu3jQ9QbaDB7YRzAfJgHsC8eei2ymj3qudxvHETcmVAwhim+ASEIgmlnOhghgPYaaixahk6G06dAjdOYxhjLNt6hjfFbr9l5TNM0onEbeKkxgzEyCHK/KJgwm8C212+T28AYMRo0CNcJjcDLi8hUL3UxRfgYaxj/6jOL9z0q2rFBm9OeFNoJn1o6TLvF+I4GGJ8d/1b5ZiUjvCOMe1sBxXNhzizP/MAe5FhtGPvV+K7OrelbiieYKw74Bd6vOTfH1H0eJ+BnpkM9DxNSvcMRhKYW9RUPhIYUY8sLhtTm+QY+5PR5gp2VcMA2xqoJdXc6apDuerMCyYiaVpv+5X/2Cj6A+WA36H/vd3DdojH4TcduYhwutVGMLQd8kudCa1wysyp+yHwZvAxjW80f7rvv06DVQDvgVzBvRmUTGPLpumF/VjfluSBwlQhwWKhdph9//JHDTmUIC7XLZs5X96dX2cODx9mb07zYh2dH2Z2jp9k7df0xe368nz1+y78/ZS8fHGYPX33MM358nh3ffZS95qvT19mjO3enz07fPMr2j1+gdJZ9eH4/O37yNlPVcb67e9mDV7g6fZk92Ee7eW3Z6ZvH2dGj12hllj69fpQd3H+efVBlQefdh6q9T68eZocPX+b3FY13s0eqok/Zi/t3soOTN6qeU5S/e+cwe/JWXWWvH93N2+asm9L7p9nh/RcFLZv6/SZ7fPgwe6WI/pA9OzrKnn3Az3dPivIfs1ePjrIHL/gmEtcLeh6/fJ/jwfmA9fscnOzN4/3sSFUA5J8fZXv7wFfVfTr37BTvYS+7+6Ds/7vsyeFB9uQd5wM9B6CneJeqvWk/VLXTdG3YgoKPLzAmnr1XtPB4O8yJV9fvnx7Oxt+nl9n9vTvI+07h9Qnl9h+/ycs9P84OT94W7+j9HAbqcZGHxx6XZfyOsvvPC9xR74O7D7KXjO3bk2z/7v3sef4S8OoOs8OnOW2zmvgX43ycPf+weJffzauHB9M5wu08fFnMkdM32aN9lFGX+fg7fvExO8V7uX//WbauleXaL+P6+++/z3799dfLqOoz6uB5irlYdhoYHGFuF0hhej/EvFWMZsPY/6Te41p8l+fWlNL3mJeYc29yzsLjK59j77KnmK9P3uaThfnP/tGznKcUY+LFB37G85rn24v82Ydn2dHhky3e3dxcBD+8++BVPudR4/unx+AJec95TM+P/48Y10drx98phipof5rPBQCR7Rdzu5Ifgt8+vLM/5cvvnhxkewfgKarpt9nJQTk2FejZ8SF40RQ3+SEInA8BlmtYvkFYqIz+/h9X/373B/X8c9KtPCct9UfkmwiUnqsBoC2Iyem258yR9UL1rpEz8KgxNaUBbmjTdJaiI2h9jA6NG3UlU9f4f/U6NHMBdYfQmEysXMtQg8YD2oQa2zpr/DwhtaGoBq0VVAzNfqHNU7UkNEF7VmeQt1GzqD1Eeaz9hv2I3GHp15LT56lVIPu72dQd2UoryGHGa06XulZ+RSlMgZrq6MaU+B4RfI9yo8yGfscw3aJOV1WpU3M4JHSREphKsfymoAOtmTGgiYubSEngU9LsQxtp5HgAM8/TKCeJK6mRruwcudbG7o0ptwzVoDnTqFbjPNDSwTzSmcz1HxVE3IAyJUPbVHRxsR+coUzXhy3B5NUd1EG/oYipY5zEQaHBwDsP/IQavWL88dK+3qQuNG7cJX7FWmEHqjfH5AGT/B3hPvBHVbMU9qEFhWZ2grLQXPR9C+0WuGs66RqPDujdoJWsQXPXMnLQYC1WNK0kuAEEUYgxjsDf6qFGTYzPjoUxFxQ+lsmE+kEDWsKiPJue8W7Ua0MPbGgzWoMedeBw5A6hvV5p5Cu+AWzGMeZm0WmeC6nZnZqOAy8sfDg3jP0N+K7MrRJKaMAGNWjQ7Xyk6I0+9cB70kmXhkYXmr3ivdcxfhTH4GkETbWL8aMXEwlvqg1tvM518pjE+CnHXeUbm5+LWp3qKXgdMteTMXXHmL9ePkYCP4ImvBwJ6Lsfk+muGRmYN4MxDuANCo2t2aJBl3loRP0qfshmeFgrSr7MpFvwf3Omwxt0lR3B2I7q0GxXdkgeCALXj8CtFNKCsQ9T0SAXGpSQA/8Iu2BOymyHnZ56/nJqMFONehP4W0UUgwlFxlB9aBIfQgl2JhXZwOTgi8Hmn3CID6lDnSlHiyiEv1tLZeQPWC6kpUGPBnX4KM3zJvbVCGCWKn01wD4s5i7xkCZJg/plY2BSYaTD/IoqQ59CHWaLor2IGT9MA3lvwnzDgsltb07MOM0p48TntaLfpDeoDWGsARNObwSvPiPvgId2Ke5SR0efegVTZfJw34Lps2T9/COCCWjkAcs4BvkQ+JSElb8Hd8pNU4qiFH1B/6MxBVqT2nrZh4gCvCMHTacTmJ8h+JWPlvsx7fU1YhsNuhQ2seuteEc1/kqg7/wB02AsVuOvfOf4kEcWzIIFYHl/8t7V2MEaAo+HdxXFIYRlh8bTL0xAvRYEXAhRvB5IJ2MKYUaySgASjBmM3CYENV6U2K3yCQuJkK+70wE7hUy5Adh9CrEjdfr++CkL6hpMTyiSjiE44mPXcrw8D3x8gtSmSdlX9EVvdmCJxoLkln0NU28CgcGlcvrxXDB40KoEnODDabc3j/3Uq8Z3ZW4VNUce3oUznM6JmgHzJp752JVrwk1g+i7x3mLMZx1i2gRjwlKbqJDAA73Epk5BarokXBbNrPyTYrxM5yIEQA1O+TzGE944gd2t+ZjGJi1sDrK7JRUY/4EOc/xKdWAJ4LtzvI00EyZL5NvAD2Pmy+BNuqqOMcaCs1UMvGLcKsj5KS9WYPZfM/LzDPJfQeAGIHALfdIwccMa/HCKqakmLnx2Sp4BQSzU4cvAk3jUxHwfQxjDrsf+kLpYmRp2zrnUarD4rbRA0IaY2OqmVrcQXKYTP8zry0tBA8ZCGj7Q415ATfitLTAIrAIDbnvKRfMRwswvgh9awTNzJlr4f6n24G+Us6EEDBAf3NJ5nzUh0Mad7cufM85yd+umfivt2QjCZCuFT0npT5fvjjWgHUz1+RV3vmPNLsFlvyyU6ULAanYgpA7ATMF41UYBfg/wpZrRynVCW4MXwf2P0UfOplIEwRQbMbjaXBtRIrPYjzK7+ve6sIUWoYOPYzhoTI+wMFvwUSw+YKovGH/lO2eNRh3+QvkQ4P7wBwxX4QD+RW3y6w1qdfs0bCMPMMl7jvHXgeYEH+Dyw7qy05bHEHx6TCUUzn8UC0F+OrhmqIUYTDNaZvfV+0BdXIQdxe2+R95kQhP+g/AdQ6PH84fnhd8fUIRBfht9f6KQnegLTSYWVko4mc4FCEKYw2qhtWHsV+O7VN/09QBzCFyrfoScX8vHUpGXx4iuxhrPNV3NNZVYOFdjpbjEO57xuuLmmn8W5iI0aUpIgyawx1q0UkIHT1rYFcw8CiKkUvwvJSVErdm0Us0Pc42kaet5TSxsln6yuJPCWjDjI4xTBOV/2cvl1uVaELgZCNxKIS2M6zlz5ImrBABzKgDkJgkTwlNA/W5MLdYWgdHqdUxqJQCxOJQLA/ZUmpp9+CJo3HTFefMUQaPBy8ScB+WatBCO9wOtAy3U4iBIwdRjCDl56ZRClGWrWBJFWPHN6uTVdWjBgRzFeTVtToVFMKWoMENx1cWHec33d7FhZpwpnNGVpLeh33DinYwDrI41MrEyHjSg2YFwmu+OhWkWzroatEYT3FJJ7VjDh6hYyCbjHswt0KK1sdECJrwaTMa8Q5Dp4/cQaGx+KVI4UQIJ5GL0EaYYY/qEYmgoUmgLYTRUK+XpjtKFfpQVFaRcE7YBNF8JDkWLIayXR1gEeP8WNFux6rc/N/74w5EL+4pq1R8IsRh7k16Pap0R9djRH9jFKKc0t1yH36GWD6frdvmmcy1kXc9HHVfF2mMdpsea+ijinZSPICSu33mcYPgwLboiZT7xmNPVFz2hEDsX9Ll2In7H+W4GbDBpUzuEU3jfIn/sKbPabUpRGBemfPR6aS5gcsNdIl8MVo/9Dfgu1zcFFhrWaMbfKIYlAMIIVjYUxuAt02kUQaCuQfDHDeTxISiVY4KFI226UNj2aJCluchmb56f0KIlbme2CGZzZNFvRTLzqDn+Oz8+1E7R6fiLyMMY5jlTzQ8jzDFYH0q+zG3pM2GT65sKmzCljmG1aE5X5/Mty29B4OYgcAuFND75G6u84iOlPjhzunZeDSqhBzvjwhh+UfwVRIlo1KaeDw0Efwf54zknfDCTYy1QPt/5eIVCSomwOwofaAO7o/KE+vDBHXcDcgotWsqrzfmvF1QO6hKHPbaxyy9R7aNJMHx1H8y5D1+2RodNmksMlE2fc0d9MFPKP6Z5HZX/VeUKQXVTv2F27PQ8ZcJgQTXA6tvJpaic8epN+Cv58MkL86aWVuQxhE0GXnUpQb72CCvlXIBlDVIS54ILvhrAaExGu6kYfW5qLvqf4EiDQY3aLOEurZSV6bfsB5q4dmyx27Q7hl/Ysp0PGGjQpvIpHIsaL4yrQnuYQ8wfNNZY4Vy1KAV0+WBIcSZcux9DG6yjk/ARawXw92orYXeWUpjnZ+Ow40GIbuCDrD6KuWDMeXNT1jq/HDY/55rMhWpBC17VVFuLVijE2FQpxvjo+vBRBJ343eok1IG21LQdMthsNz/OFyv9Cq8wj+G7qdXLCczvuwCAdz9D4D577DPnqcB3YW4xz8lNiwpI5iE5s4CJs0XDsKABczuavqoODevY3aum0fJCAZq4UjhiYZC1UfoZr2h5LjKPrcEXc2TMtGioYmFXMF/DTUQDfywoXJmzZUSYBLud2/BrLPOt5Yfgyx6Wrvlik9k0D1SrWCAvHhGUeiO167Z0QTijd/JYELg2BG6hkIZpDq/m3LGZV3+wZk517XMrMQhdbRcfGZjnGk04rCYwVWJVpnyteYWGDx3rEjjNMzkTq0Z9jC33jSa5A2gZ8CGdymjIq7QbMFu1wBx5dTuAj0prkjPvmgPNQ9LLj0zA9vUmVqE22quzk3fcxtELNjnYFJCwRoVXi2o1bUwZaG76LD+4+Jjydni1PE7JQ4MWnG3XJg3HNgSon0//3tRv+Ie06kNq4qPLprewDSFUZ8YLsxiELYigED6blOJYEu6SMldMV+QEZ+EWGfBvspwmzJ3QBGFZrwQNfIg8rOrdJnybgLdtuzSGz2CfT6eFidSDPx5c6dXxI3ZzQDpMz0ruWVopQ90268e1YwutGLRfaWtOi1CCz1oGmIZjCOi5CYrfERJrteZNn1NzY50aOGIjgB+Zg0NRYS2FRgTHCWAMhdBeDrAoGOG5g2NbGn0MaLwJp9sl6mPhwJi1IcTBD5A/SMvnWPGiZK0gH0HjxrQUpJWks6bVi3BMiBq/GhYLHYING4IYvxv4HQ3wbjRsdHGxCaU3wKYbZMPCwalDc+FzHwdk619GdIhZny/yixlFuchjDBxo5HFkBnz0mq0hJdAam4oxbBj7lfguzS2Y1PlIlfx0ChM+niY27liYR/BVrUGzqXyy4GPWN3BcDd/HMTtjHNOCjUj8JIKmmscAU8yLLzZ9TtetSvNVaKM2vbvluQjaeaOKsTD+2S1kTgBkhOC7FmP8WDi2BUvhBX5otbtUH8JiAB7SGFlwj2gogauSHy5ZDpQ5c8706cPKoBbZWGiMcdaHA41+3mfVcUmCwI1E4BveGrpLyk5OTuinn34ibFWle/fu7bKpLesOoO2Bf4+Pj8WWJS4vW0xDCG8xtFGl79Dl1b2hJuz0a9g4h8jvz8wOG7LLo3MgINieAyyIeDjTr2OMIUSweLC79MMPP9DPP/9M33777e4aOaNmFpS8Nvwwy9XcGflv+uNzvTv4orkOwi55N5Dn4Hw6p1OnEXyOV9YgN/0lCH03CoFffvmFvvvuO8IRHES/d1Zp+/Mfif76ns/gWH225Z1buLuT/cIKk+KWIF1WttSHL1qtvbij87Iq31AP70Q1el0R0DZgdNFHgu15kIOJPMWu4B0LaOehaJd5OQpGaeHcZTtXU/f53l047FE83dF5NRRu1wo03P0JmXAlEQFtO8Qk1/UicAs1adcFOLRoMPPF/SvWol1Xd6VdQeAaEbgJmrRr7P71Ns3mVxzJ0r6JWrTrRUZa/8oQEE3aV/VC6wiT5H1VPZLOCAKCgCCwgoDWoJHfWLktNwQBQeD8CPzN+YtICUFAEBAEBAFBQBAQBASBXSMgQtquEZb6BQFBQBAQBAQBQUAQuAACIqRdADQpIggIAoKAICAICAKCwK4RECFt1wh/6fUjSDKOPJMkCAgCgoAgIAgIAleMwC0W0nDyfxMhhaoOeL3iF3FTmwv7iCLAh5BKEgQEgd0jEI4QrB5huzhWr4GA4u0xjnjdMoU9HD7cJJx1fPE0aZLe8i5e/rwlgw6Cvo/yKCaXQf952/+s/D4O89YJR+FJEgR2hsAtFtJwKjtOZ5+09AuCi7iV7TYhUsnVpAAxRPvXwQ0Sijl+kSRBQBDYMQIhzpBDjNYu4tgiWkcY4iiLENFJhgWTQfSNnjtATICKZLRoAj5hVzwub0cjlzpFlJMzsl7t4y3pv1qiNrVmUtdHeLubdljxWeNkU5fk2Y1D4BYLaXlIkiIc4vlfDCbCEIwOkY2uJIWTIcUIJXMdKf16TuS8DvikTUFgOwQQ05NDFznTMHU6NRC1IA6ivDyee/hdvWTCQd0IO3dWqKMYgkVYXcl2tO4k13b076TpC1XKsZg5RumFCu+u0JnjZHdNS82Xj8DtFdIQq3KqZvcQU9LpIOixi/BJBhl2hxD2Mk/xhNoOgvYimJ1hNmkUIZzkpANG2qIxgoR37CYhDJwKGN5vIgadhTh3OgJTd8pA5PyojxApiK2IwHGm08ep63nV0biNe1y3gRifTRqibpXYBOAg5h6CsFuNLgJ0I2YjgqqHgybMHwi6XmQr/6mqn0B7BzTZiH1nmojj5xUl0V8L8UB7LupDv2ycvh0ggHGzgRiMiNPXmXYefUX0d5HRlgCXS0FgWwTwwRxC0NLxJdcMBxqsqLpkDfGBa4iNWjII5NQQDzXoW+AJfWo0EFsX5lAXMTk7HK1+4pKBWLbDFuatC5NhMiTH7CC6ap5i8KmGjfmPOW41h9DAIWZl26YWmJgH3mKre2sSeFmviecIsm5YiJBSKPICxLJz+iPqwBzbGHA/YlWfiSDmFuKQWi7qU7wNofcQ27gz6JDLPAW/W3lgUTxLFR8zYcq17QaCps8JnfP0x9AImoiZ3ANPBv81EQx9OCU2ojH6bKBeB/GABx3wXfDHlZQiHB60c0O0Z6GuCbcesTk5j1+q2p/SFSH+bU6r00A83Lk6ud+NUcG0geEQ/e+o5iLqI55y6TGzgg9iKw/cIn4u3kMD4QBzDsyhCZvU63Pf+HmLRog7qjBnftwaA9k8VdJbhc+6ccJ04Bti8reGvwP+8hdkBTm5cZMQ4Nidu0w//vgjB63KELtzl82cu+6Pz4+zw6fvVbkPz46yO0dPs3enfPUxe368nz1+y78/ZS8fHGaP36gH2duTg+z+i0/qN5c/KsqX+R6++qie4WF2fPdR9lr9fpHdP3yY5Y+47oPshOt+9yQ7RJvvVdWn2ZvH+9nRsw98kX16cZzt7T/MXrzL28qyN9njg4fZ65wMlWeaqurP3mVPj46yJ2/zQqdvHmf7R8+yDyj4/unhrL+fXmb39+5kx8/egQpu+362//jNtHqm68GrdQ3P0SA/BYEbhsD333+f/frrr9dM1Sl4xr7if9O/vcPs2ftqsj69OckO98FzXrwD91lKn5iXPMHMLtLrR9ndO8fZ07dFzoXn4Bn7x1nOUk6z148OsofFPGY+VjmnXz/M9u7cz1584DZOs3fgFfs5MwTLOsjuHJ5krws29+nVg2z/wcuCzo/Zi/t3izbeZU8O7oCfFR19/zQ7OijoVvzqJMvZ0ifQtZ/t3S/qmKefeege88qc93xkvvTodc6jXqJdlFFknL7NnhzuTXnnAmKnzNvuZvefl1gyjUfTOhVv3gefRhOnrx5m+8fPFX8EhwTvZPrzK+73g5clD/yEOg6zJ6prH7JnRyXGzNIX8Xn/9Ah8Na8jO+U6D7In6uXl+JTP+FtyB5g/V1k/5N8Ila+aXvWNqcAHTHxhnJwyXgV26ruDfhavkBuR9BkIsFyj5va9v8vo7/9x9e93f1DPPyfdUk1aSoEfY4WnAz/89mJyutCsTdXWWNHWWZSOsHIl/M4fmJ0xDRoc8a0oDw1YnjRyBh6Cp6tCeYI2TcevaDSgGFq6/BHydfvU4mJGhzyvQ4aqmv9TI72wnQZeBM0ZfNBMbosrCSioY5U3pS+/rR5V1J9O+jQ0oIWz8kK1uk510J1iLRdgJdXolf3FCrHepG7LnJpJtKkNN6EoAl1F/2etyi9BQBA4G4GQRuO/LGb77U802qBN0+w++V6XasMGtOmdqRarqq2a3ab21Dw6l4t5hgaNkc73atCWDwhTfKtUg9aqUZQzYQHQwmCq2TFbsCIUbE5zhhRMg5TXoaWpQ5tXaGlqBjTzBX+Edt4EL425dWiMIqtRxBEGP2w54EsVSXeoWfCvOjRMWhwrU2/ohdAcFuVqFkJQlXx4tZ5avUEddFxxUrioTCJoAYs6CTy1ocNCAcICD/GNW9g0oaqA5iwHYLXCM+7M8InxHlNynLxGAh6OXYOvYanFMqnZzJ/VsUlEg5U3EKfYAAAgAElEQVQmb1KHlhE6yggNbaBXkVGBzwqJGnqPChX+0LSOBw2JW7oC0s29cQsDrPPLgLo/0MnusQDjkxdi4mICqQTVsB8b5Op8YVKrXcfkgl/IYER9MB11G0YCLg++MU01TKhRb0I+fEZiMMjIGGKqQzUOW4HVKTNC4IGpokzRqE8jMIcIzCf0NXI7TAP8UkIwiV4hoOFOCsaWYnKtMrOEvIr6/bEPc+5g5p8ShRTrDdAfUp/7WzKqwAfThFm16H7gRyinFyQyLegnmIYkQUAQOC8CdTL0PaK//DZX8M50MVZVWw3ztO/Z5A5a1IRJr+15+cJuTYEaPsDF1F18mkKggfm05CI1A2bPNeXX3apjkTblPhp+w9+BhSNOGn/wy1TD7wB8cYSNDlGMtWRE1Ckfasg7zah+cB0phDimeZo0LB7XdgA5uO1Zxmk/4YGBR3NPKivgOkDHtA4sUdkFBDw49yNLKQqxYEV9ScyL8bk6Gdey0zNqz/w1wwdtJRFM3Q55qlpchxHVOkWltRrh/9NUY3N4eTW9X02vAqMCn2Uia06X+uMGzMdwzxlASWCtfkmWy8j1zUHgdmrS4HMRQADDwg8rDF5xwi+gnBgQsEIdWqviHRkt9tdqUtiGX1jpNJagDFheuYCNsZq03DHBmYA6/SF1bQ1+bSzZQOCDQAQZbzHx7hv4KHQhDDY7PRoMXPiswG9ERzbU7acFbUWpACtHwy4pmq+qon7WAAY4XmRO9RZCGNRBSC2C0zD6Wz6KsIuszvdVtRDKlPBZgBF65NfhJzHjXYv9kCtBQBDYgADi9fZ7dHxnlmX/wYB6jbmv83xpCFZRlBR3NDJbI/h8JfBd8ja0UfGIZYGKZipKzN1eKlhRTwK+aOG4jjp8uLrge33X2KrJ5Y1IF5CFFrpwHp/ZGqwGA29Ckwn/edCiYUctWPVMDD0bna1z1LDIH83a8vFuvdb5BKQqeremQWXEBpQBvhV9E26MONal/I6drxLJfU0I3E4hjbVHWFmyGJX6HsVQN+vFC0jwLGX1M5sFxxPlCKvBeXYM591gEua5WGU/LRNQvxtjMsI8yRsA6ikEHejgWAIEAwhrOpyB82JJMKYJKkzG2BRgQIvWdrDJoA7BCcyioIfNAdPfqlgMFTmbEtZwyor6WVMYxmh3yg8imng4csSGwRP9m/UXtHpJTis3VQiI5UIrmkDwbDhTbAqI5B9BQBDYEoGa2SYvek9vXr2kV28/UAjBRq8qGw0x3XpY2s2Sbuhq8865E7QsNaiHypJJAGEh9+o/s6qkMCuqjKgjWdDYzIr70KCZnT5Mksz3sJIrTZ0bWmAtGmvTpglmuPN2jxVxSTyrg+ndLkF7BUF4lhtCU7+vTMo1qM3mjxpis+0Mdea9530HaAstzZFJEfj+YG5T1tk0V9N7dtk5iP0RxiC8WmAaH0/alPQHC2PsPHVJ3qtH4FYKaSGkqDo0Uzz1Qjid6XN2S9ZamUoLFtKgXe7EzP3WrIal3lAUhPAhKFaNKQtEUFUrGQrq81Ebu2cKEyGbCcplHnZ5DVrMgDF1o0ip4VURaPW67RHV2OdC1T2jjR+z+dWLsMtKV1d8A8ykYCAV9efZIphR8xIxdm4O623CQhf+HPC9mGrlWBOHnahmUTU0ZyH8R7j3LOiNR0RuM78qcsg/goAgcF4ENMyxBnyOoCpXc74qQYNvYw/icLr7DmcxjsCPsJtQJTaR8ZyvKj9/vw5tfOLBB4tvYjHWwy7IMM/AZrZNgl/iDanc9BhC21QDg2DetJw0mBmjkskkExrwTs2ziIP51vDH5Kl8CU2GE4gyZxVabNkEPcGwOOSX253u0FymcOkami1bh+/wuBDw4jG0lAF4MbzQIGj6o3JXZUiTcQEWqqgbdcU3FZXRmEZbCVpYWONzMR5i0c/lwP9HvSG+KhtHwCLBG+jd2NOlcZLidIHOOJoVmTODb6xHHt4IBG6hT1ruOG926ngBrKWCldItWVCUa626PJHgl9ZKsV3boSFWWTWri23MeT52wo/bMHdGA6zE4DOCAyhdmC9HkKQMaKt4A4DaEKDjtP6GC8deG/y1RgaEsQHLPNg2boBpWwE0eDq2RXN+JYWxMJhidcq0FYm35WsTCFgN6o7H5EYtPDfgE4dNB1X1g9V3+uy4a9EY0mMN+YbDJnzasGWezaCtor8wfbKpt13wjUgJr92cjeBQzJEOx2UmS5IgIAjsHgE4wfewyGuBtxiQI3ha1m1oX9yCH+B5A5uBmlhINWAC7W+iqAZ3igGOvnBtGnNoN6NHo8LMajgNSmD2MrF4G43Bi+bqYeFNB98I4NA/YI0XzIN9xbRWkwUtmtnEJoAx/Mrq8KPFMQ/BWdo01NfDMUJt8FXWqpmgxcLxFucR02qNPjRS8Nez4HNbx8aBhgnhZ5ukw8+vSz4wMfvsc1Yja4BjSxho8PDO2IX/7UBFe7DxV6Y6jkFyRzi2SPF2WD/wb/nF2NSq2R1Sq9mGsMb+ZilpjSENIbhtnzbQu6mS5XGCvtnNJtkT1mLi84NjVPRN5eXZjULgG94aukuKTk5O6KeffiJsVaV79+7tsimp+9IQSOG7AIfXNs7umfGqS6tdKhIEdo3ADz/8QD///DN9++23u25K6r9qBNhkmdap9POP4Iri1qCdO6e/1yayYwi4jXRIflvflE2e3XIEfvnlF/ruu+8IR3AQ/d5ZRePPfyT663s+g2P12ZZ3bqW5c0tsbm+2eETDtEVtEdBu7xiQngsCNxSBFH5VFg58hVIIKVF+tbqxjW7rhnZIyBIENiBwC82dG9CQRzkCdZdGIwFDEBAEBIGbhwAfKdFl06TZgzYN7hwmzoS0C5+Nm0euUCQIfBYCIqR9FnxSWBAQBAQBQeBKEcDBsC4c8t0dNlrH0SL+DuuXqgWBbREQc+e2SEk+QUAQEAQEAUFAEBAErhABEdKuEGxpShAQBAQBQUAQEAQEgW0RECFtW6QknyAgCAgCgoAgIAgIAleIgAhpW4IdTxAkHQfO8hk6Bg5kdIfh1mf7JEOcCYRgxPlupC0bXMrmt3EO0fg8pwldrJ2yFG9Bt/qRurwM+j+PmvOW9qmt69QWp5LzAif5rxOBhM/s0rBTEXzGwCn+OoJy94rDUC9E1+Z58EXM67BHptbEEb8b0jZ5NhRf9+iLwGaB8M3vel0f5d6XgYAIadu8p3hIbjsid4yYnTj9NsQW8LTr0iAqCocICQVmWpW0Jk6axmGymzeJ4/TtdptGcVUt13d/O/qvj77Vlk3q+ojkcNOOEDljnKz2Q+7cOgRw2OsoBJ8JQ0Qf6VMdQdb7253UugaqpXmwNP5u5LzmuMbuYHY4rdGiCQ7Wttf0bnprmzybyvOzpXZvJDYb+3BDed7y+9zYB3m4DgER0tahsnyPwyXhpGkcNJ2negORCBC5oIyFh5P7/WiDloujBiCEyuaUIERV8Fnats31f8bTrej/jPovvShO+K7jVPCzIL/0ds+o8KxxckZxeXzLENAscoyYwuii/V6aB8vj7ybOa4RP8gKEl5p2WaO6XoTQq4RhmzyVhfMHy+3eRGw2duGG8rxlXDf2QR6uQ+D2CmmIwdZGjE6NzQt2C6vXDUIWQp5oENSmIfWApI1AeEPEE4mGLtmdCUK6tcl2WirmXYCwTQ5Cb3QQc6/B6raJS7rrFfjj8MU+QnTYiNNncngXxIRLcbK/g/ArWEH3GjjpmuPJrSSETvGhsXNghkQIKdMdIbomp4j6qKsz6sEca1OPiyJmXw9tm6jfhOlEtcFZY6xITcTw67kI5Ixnlkuw2hYpopGLuJ0oYzsu9ec7O0d/Om6gL33qt5rksPkXIVqCErp4ovpsWg41sBruNQxqrVMwoj6jOaBhC+FZ0A82A8fQTjoIdqdwAY4IWZgnVSfTijpb83Xm/S4sssg7QbicFnmqFJ4Bi/zZGnwQD3XgAkduC3+NnpcLx1EfbXcQY7WBYPQI1wUa/RJz4NIcRgVR1fRW4bNunCDQK7mMIfptIvgxB3qW9JUhgI/UEPNKx4pBMxzq5ME0t+4kgglxmF/FQ4wpD8EIxyn79oAHTEAduF90Bh1ywTtM/G5NY1nO5sHa8VfO6xRzB6Hj5r0p/JZJ7oQnNvhVD/wK84nnioM5MZ3v871gU60F3tIBX8DcMa0mDWaMgcYIdcX8yELMY8uFZUHxjKW5+X/3qdHoIZ4w5oUNrBAiD74W5JidIiD4Gt7J1cznuQgd0FiutLvAszHfPXZ3Af38Bxpxfq5KXksH/xhSG+GxOJSU0yl4Sf549t90hJBeLRoOEA4QPHiCJ5U8D7iMmDfiXTqNNjAFr+rlDJG/LQ2E0cpTgm+QSR31aJ7nrfkGVfE8Hj94V70+fxOYJ4L3Bvh+NKGUAG+ycWhwyZZS8KsWeLHi0XaD2uU4q/qurMOV6UD4MOZ5lukgzvXnOAEtg/yVXnNYqF2mH3/8keMhZAgLtctmzln3++zp4Z6ia/p3cJK9Pa2q5jR7+/Q42z94mD1/+2k10+tH2cHjN9P7754cZHcOT7LXH4tbeL7/qHj+/kl2ePgke8dtnb7NTg6Ps+cq34fs2VH5e7WJtyf72d7h0+y9evQhe368nz18xZVwub1s/+HL7ENB/5vH+9nx8w95JZ9eZw/3j7MX3MbH59nx3kF2UnT044v7oOt1pkh5/TA7wG/VO9D15GAvO3pW1DFH/+mrB9ne/qPstcr4KXv1cD978JIvTrPXj/ZRJqcwe/8sO76zn52843xLCfXdvXOcPS2x/PQqe3hY1olaXz3M9h+8VLS8RV8OnxaVfHiR3b8D+tUl+n18nJUkZhn6efA4e6OayrHMn63i8/7pUXZcFjzFWDg6yJ5wnR+eZUd7h/jNiKBvD+5me0dP83f16WX2gPutHlXTW40Pyi2NE35POXb8ao7RzwI71QdJn4PA999/n/3666+fU8UllAXfwLxd4DMYX+UUWWngE8b3/uPsbfHg09sn2dH+w3yu8dgpeQief3iGMawYxzvM1Ttz8+5pdnQA/qLqmJ8HuFwaf3yd86VP2csHB9mMhb3Pnhw9zJi9nL49yQ4xF3NWxv05mPGF+Q6A9uO9OxjPec5PXHfBrz6BZ5TzGSM9e3H/biXvyhgD5o9l3fPXVbxzPs+F6EBjy+3O8+zT19mjg/s5D2VUwTeZVzIrePP4bnanxAd88/HB0RxPmgPo9GV2f+9udv/5u5zHbuQh4H/Hz/H2ODF/4vebX/G35cHLgtGjphf3wa8U21h818vfoEqeV4yfkh8yH7pz536Wfz74O1PyW35vR9NvR/aRx2rBDzd8V5ZxPX2JsVBgl73DtxD9LD+Tc2h9MT9ZrlHz+97fZfT3/7j697s/qOefk/7mK5U9N3crmtD4T78t5vm38Ux7s1IagXixQvIHJo2bWAm2x8VKcCXj9IbZ6pBTX32eBiElWEGYbIpDINx2v0PbHpZtus0iGDJWb00d5tF8dcXB4JtYweqFec+COm2MFZ5KGlZDekJxuWCB2baJgL+c6lgpaXGstGyhF0IrZud+c6DLdY28/Jr/alh5Oby6R27LRrD5iNdaeXD6JrRnKhlNaqpOrk81aI7alqoEjU8oQAB5u7jUEHTZDH34pUTkB3Vym2aeD7QjlvIF0jw+McLIpOQ4el4PDsZ08AKCIAeoBm1HTrdGhqWT0QAt3A3gaNVjijhbJb15levxWSVbq9UUdox/HX6LI7xTSV8TAiECmP9lsUO//YlGm7RpCMnW5I0D0Mo0ugkCew+KubYBF4zh2byDlgLzhmfk9kkju6GTPwnyIuCPvgZNCsZ95HlUh4YlZ2Xggw2L4infWWyhpkG70sxzaja0b7GvtG4aNk0Fo9Intw4NSp2SaSD2Rd61ieZteefF6NjQMua7D01jo+DnOqweOrRNOfdlTJwcnxo0aUZCih2uSTW4yXSgoVRsbgMPCbyADFgqdFUHNGd4NxdJs2/QZp5HGDHNgvfUMe40E31VTcJiA3YeR/gJ37JJhDFZfDuoDguEHlJY9rXiu7JCN6uFUaEqBg3peNA4w1d7pYZbd+N2RhzQdKrfwbv+z7n3vYcPcjEJq0ZB3e7QJGzQqA0GZIc0gQq8Sg7RlI1iNTFzqs0906HW3y7VsNtrRiDXn6b8eUeCybYMNpxfwgw7gWlw7INhxPCdo9np3Bp8taYNznw9kjhFNSyN5Il9umq53DK9N3s21zcIGnlKYRipE/hvWUNuplkpnd9gDKatJSjrdSE4DfJ7CKAcJBaEFxaK6jNfQFDO/cyZY0XF624v4IO2kggmKASQV93AdRhRrVNiOcOHu8bm8JVUSW+eU1t4GTNMl+sxOwOyYVYwfZg0BjCX68s55PrLRqBOhr5H9Jf5BeGdhXm80j+1cWBA1sqDTTcKk6jKko+3YjRvKrTwrA4+VBt4EO/wkfY9go+E4hMRxnqATU/OuKg3xiJTb6iFxcrIZj/Qslb4dGkaXDQ4I/O7YET9kUcBJJgIPmfUKTIu8a4FopYuKnnnMp+6CB2bGsYKN52vE/3Rkjh3kVDdK3sNRPD/Suy5XNnOBh4C9gcf5hnf4fprlZVWEz6j6yyeh9Bacy+TvwOz91jWjzrY9QTjJP9MpBSFCYLQ4zmXrfiuLFPHIb36cJmxTR8m+j61rPpyFrleQuCWCmkN6vcfkP/f/5n+QwFyh457fXKrxgs+6mEKhsvO/7xqHUwosbGbBn5U48YKqzp7kF2gSF7pYsHa/MyatpqS14aGLsausG5XCZ7jZnPKUKqJKxhqmeECTGGROy3VV92weqI58PMYOku5fBqlS2BdGLu5qrHibQ0nOKZjqblo6XrD5Xp6NzDodXVhNdrDqrkBAa0FHw137CGo/WV0cF1jcu/qEYAWuN+jkf0/6F+KBeH+A/arvIHvWIefUdqBr1WL6pMEvlX6FC6rC8tD8+I0J2P4YfU06vbb1MVCk4+3KGW0c7+Ti5NBl0rHPOEXpKmKh4yrxbxzwzUtUMXzzqFzrWEBMfBwJMoyFfHyjU3XsAINArI8CGguLFOwUE1cfVOBW//sdpo78dp1FybLD2/o1ctX9OZ9BMHGXF0ZFsMjRjBfu+PPDRYIbEaNUqyGzptYsEqnqv4UWu8xtEbb1AK9ErRiZeJV5YLGZvoEps5JndrYum9jgwEvyKYatw3NaFgeJVObKHxxCzPchiJLj1jLBbPqlMR47vcZtRSYTNGEan3QGWNVzys81DPFJ4EWbFYXkJxd8JJ9q9eBOsGY5rqKTR89GvhbFc7bq6T3jH4uPMbu4PEEwj9Mxq0hTboa9QfzY+w8dUnem4pAzWyTF72nN69e0qu3HyiEwKJfhFglCMzGaKlEv0hV68vAlGXHNJl4NI54V2mei7Uq83yH4NTfGQbrq8Dkn01P/MbYZv7jQ4Nmdvpws8DZb3xjfhKvr2nt3a1552XTAZpr83WC/mRBc7SW3M03N/CQGtRmMSwbZWJeP7vigXAOXqUquQSex3VAxTf3BSKv3z/3ZqfYH2E+wL0D7i5jKDmSPjaibEbq1j+9tUIav3kNq8cG78oxwDg2pLqFHY3Y5TfdNIXdkyPYymxrVo4Ftm2mjgabf82DDxa3xztd2n2ayQeslq4mxB+x4MIJzHScYIeMviYzM0b4hxW7VXlH6BANnCVPGtAMhhAaVPOgazgJ19S96ZaOnU8xjUb5TtLEh49VQcOmUuoZdjGZAfJHec5kgt2jAczC+JxZBnbHlQ+gbh9N5Zg61eEj5iEfp2SCyb/NC4BJFkorGiNAs8qO3Xcj7M4KK0X0nKaF/1bSuybv0q35cRJAg4ZFpUr8AVprWj27Sslx0xHQDPh8wc8H81XJWhdJ2GFe4zMauSzG7BhnNl4kbeJTJnYKBv0u/ENhgi8q1+G3Go9GhYtBCr9c3tm4vhdpMqEBtHCcEm8EPy5b+dBqsEBEWPTlD5AH87lSyMQ8qOHhuqm8mXfO0LgQHRvaJeBh4TSASdGFGLwmAg/gvl04beAhBoRZ5vV5c3CrGc94cR2mkRDad4UPaBpttbi8BJ4HTZytezTEd0eleIwdoRiD2hkILOGa+n3qjKNZITaLn1HFbX98O82d533rvHW6G2H7sQE1Pa9KoP3ojghBAPIEIc5uuzi+AluKJ8PNkxfOkkO3heM/bHAybLBvw/xR52p0+CfVsA3aoAk0K14HksRcSmH2c5wEx1E40AJBi2ZD+7OYpciNYy96DRxXga3SdR3b/hvURFNzC7OFesuLGraV9yYtrKaHYKpwVkWhyby6aW2pxZtWd0jNFnyrQBc7n7JD/jpmu1IVThTv9z0cIWDSACv3FJ+ILo42YVgavR5MLWCIIwP4wokfded1Ao8usMQza4Bz6Hi7uh6tVL3uhgk6W802hDV+k3gHjSENGcvtiqNz1fRu7O/SOGn2cGQLtto78N1JUot6o/LTuI5quXerETBc6sEZv2lOcG4YNrvgSIz1otIGlJbH31LWGo5VMEIX/pkzxlKz+zRU7Y6xiIDwpPMRPuvFkxqEMh2aQrvHQpYOnyMcPYQ2dGjRTCyGrTE0+4q3WLAe8Md+TQ+waalhdLF5go8n8nCUzhyRVbyzkBvKnOenA5xmud15bGrg66ClhU0TA5CcajZ8SJ111G8Af+nRBh6Ck9EhyLhYoA7UBhI+2qNM9UYXm0la2LAFngdnfRP/biPkVPK8rSnWyR10yXexoO6zjxw2TAw4QsYZFSzjir7ZcL+xJ+C8eG8NHFWln1HFbX/8DW8N3SUIJycn9NNPPxG2qtK9e/d22ZTUfc0IpNi1k2JTRu5Yymf42BRip2nv0mSPlEYQ5PxuSP1Lq/OaQZPmd4LADz/8QD///DN9++23O6lfKl1CgM8nw2aqXrDGZ+kqwbopdFxinzlEXyMdkr/iRHuJjUhVF0Lgl19+oe+++45wBAfR753VOv78R6K/vuczOFafbXnnVps7t8RIsm2JgN/F8RWjXEnPW7a9CFu49S0LSzZBQBAQBAQBQUAQWEBAzJ0yIC4NAbuLnWwwP5owP/IZYEYXJ3lvo4u/NAqkIkFAEBAEBAFB4OtBQIS0r+ddXn9P6g1s0W7skI4aNeFE29xhC1K1ICAIXAABzUUg9AuUu+wiN4WOS+xXvYWD1C+xPqnqy0JAzJ1f1vsSagUBQUAQEAQEAUHgliAgQtotedHSTUFAEBAEBAFBQBD4shAQIe3Lel9CrSAgCAgCgoAgIAjcEgRESNvmRU+aiM+G2J44r8bAKf46zgNqF4c2blN8JU+Ibeo4J2ey8iC/MWlqOIsmqnh6M25zaJcaAifjqJvKtE2eysIVD3ZRZ0VTl3P7jHd9OY1ILV8HAmNqIpalbuQB1k3DIKuJg5Y3Hr63oed8HAXOFrsJrmJVVPLxEtY5eF2KuI9G+5I8tL4AfKpwk/u3BwER0rZ81zWnTyGf+I1g3CEOHZ20cLDflmVXsuFw3EmA4NrFg2jkUmcy48Q2jqG/cfHMEPTZ7c16rPHHA4GXN23e3CbPCjbLNy7Q7nIV13q99K6vlZb5xpdwvTF03XZCNIcGYUABeE0QIpZr2kWc2eJYm9uOjfRfELiFCIiQdoGXXrMQJgShWaJNaqSN9SJ0k45Tm4s8se8vrJZr0NpxiLsblSKf/GhuSY9wHjrCvWxM2+TZWAEeXqTds+q80ueL7/pKm97U2DKum/LKs2tCoE5Ow6AoiK6pfWlWEBAErhuB2yukIe5Z2zZUvETdbm0fZ5LfmIotp8EEShR0EOZkVAovfMq+SZ0QeWJoykwOoeIi1JOFcEYuDfk+p6maPaFx26YWYtl5HQfhMjiGJAISI96UqtNvk2EP5oLa4kRvhHYZsXDIcT9dmApsm2z8NXreetMjTLWGi1iYLupHHD4L4Z+8UrhEDNIe6DURa8mEiaWBWGyqJ+kIoVkQCmvQJAt9+L8QosnuTBAqDiGfHIQy4oX9BCFfXI9zq1iCQ64fdFgIjdUpG5jPw322gEeHw0UxHk3ErSxwq6AjOqtd9Njvg0YE4+Q/p13GNiXyWjowGVKbY7PCTO10qvBBOK8m4psiPJPpjlAj6uwhbAnC7nB/HDwrySTEDu2od+kg7BbCeeED2lLKxYj6yDuz2kwQ1qZFCp0Fk0qerzPqIY6jjdPZAV00QrgxhPBS77FB7TJAbNQH1h0atBoKLwt0+IjD6iKKg4X31RxGXLtKMQIVO+i/qgPvZ1SMMzYN6fzuW8Ac795s9FVf1uIKOlzOg3pMBD8u4xROG5EfF0NAzQ2LdITh0BADsjOJzlVPgtBsHORcpYp5wo/iSQch3fD+eY4XfGShIfCLDp51VKzHSIUjMziUWoPnJMZULx80zM8chOrpgC80BhE3WjHHNoz5TbwPbY+Ah4ExbDsu9f2qlW7erhrTyDvlTaAoTfyVMa36WskTq+uaYjSPD9eD0FU8F5if9SppnJaWH4LA7hDgsFC7TD/++CPHQ8gQFmqXzZyz7vfZ08M9Rdf07+Ake3taUc3rh9ndh6+Kh5+yd8/vZ/v3X2Qfcefdk4Pswcuy4Kfsxf3D7Ml7PPj4PDveO8hOiko/vkCZR68zlfPTi+z+4ZPsXVHj2xPU8WrWuLrmOk/fZI/2H2TTR6jz/v3nqt33T4+y42cf8hpO0Z+jg+xJWWFRr/oHtO/tHWd51lPQe5jtP36rHr15vJ8dP/+Q5/70Onu4f5y94MpPX2b39+5m95+/yz7lT1HPo+zg8ZvySl3vP8qvPz4/zg7Kvn14lh0dPs0Ygvk83OfjvTvoFzcACLh8ka+SjqKOqnaztyfZ4dGzvC1Q+uoh45ZT/Obx3ezOg5cKq+z0bfb44KjAQDU/S6Dj7p3j7OnbvNwp11mWw9vid3FU4PwWeB0+LUD+gHd4B+9XXX7Inh2XGPM1sDx4nL3hnwvvGvmO9rL9hwz4fzMAACAASURBVC+zD+p1f8R4OZqOkewj6tx/lL3mZ4zjHsbSO75A3x7czfaOnmb55cvsQZnv06vs4SHKFC/q06uH2T7o58vTVw+yPc6nnjE++1N8lt8nv4MSO36fh0/VG/xi0/fff5/9+uuv10w/j5/9RT6Dd/qsElq8y7sPMXry9On9y+zRAY+B/Lp6nrzJHmPulnP89aOD7CEzjenY43l/lB2hIjWaXj4A/5rNjSfgheUYZ3525/Ake51P06x6jm0Y8xt43yn4EfMKNSQxL58czNpeeFnvn2SHzCOZYOQ7OTzOnoOmTWO6kidW1FWFzynjU/Kzd6DjOOe5C/TJhSAABFiuUTLEvb/L6O//cfXvd39Qzz8n/c3uxL8bXHM0ofGfflsk8N/GUw3EOspTaCt444DJmiLfpmERAHxd3uk9BMBtIog3pzpWsVocbxdwvKyAg9NaIY3V6pcX0hNKEWC9Dt2a76UIuK7nOWtY/SKYeRCsX5Vq0K5AsYRUQ6BjBGaGzwsrwyyocsb5A46ITraeIHh7USUOpu20QHN+ufG/gR+SAQ2P6qmO4OM9Z225mgZNUbOu6tIQtNmOfaXZ2UTHpoYjmIlr0JQZeY3QepkUTKCeUgkBgBuMFf+EJs1IKOJOr0k1aI7aVt7TyPOojr7kVHIdFsXoH2sf/KBOLtpQCe8WYUQvkAxqQpuoM1gInTWJoLEoxgjVoRHVQwoLOmvQvDRNzqiRYWHjSgMB5tUltCX1ODe3hxMKDGgLixelOcgTwnxeUKZBY+moZxq0rjrFFSBwhAh+xiOtDk3MqKkXNcg/F0cgpNH4L4vFf/sTjTZp05IJtdTGAQPvvEWB06N2Mc4q50kE/zUNmin1ymrQeg8IU3eaUh9+bThketA11RwNvRBa43JuQCvr5DOoLGC2OuTkEwDeBpvm2GLXFq4qeJ9qu2Hn/AH8zXUX2y7rSIOQEvBaNd6Rr93vEFicSuvHdDVP3FQX17eMjzKRIA6xmobYeDEeNNbys5wa+a8gsFsEbmfEAfb5ugNg/3MO3D18BAvGtA5ytXFg1Fj3qPqehvBI06czH7TqAstPwHAbOnVxyj7BNOvDBmWpILsRJUmkTIyeaiClJIyo1uFP7GqqgY6Cv4HD6aQpcy14H8wo0QRmu7EPASaGozKRWxYHo5rRvlrn7A7qSlBtaZKBeGOVHH65WH0OD/iraRrKgpCNdCzXMXfNpiDNmFFZY0e+ZCYIa8xsVULv8f/16HD7s3eTJikF2BDhjHPE0hgfC72BsiwU1cmejhH4mqF6QHa+BPP6zN8Q741NqDaEWtVcSlGYIJhyTjIxRkXtkKGUaX4lgd7E60JgH+TvOMV7TKxpX7V550aupCKZ+LDbTXwUYWIfDGDW1isyyu1zIFAnQ98j+sv8gvAOdodPB9FqXWrjwIgc9SSmsQuTZG+s3Coq5wneeTo/VgyYCLk45iUvBLqtFO4Tfi7w8G2Mr/lxoS35ls7mDU+nqjmmcwvVqYL3JXE6M9+itAaeUGM6l1KSsJl3Nt51zBFOPDXWj2nMgwqemIBBraurCp+a06U+XAVs06fOoE8ta8P7WiZcrgWBS0bglgppDer3H5D/3/+Z/kMBeoeOe31yb+Bc1OCnpPXhDwZpZhJh1WsWIwDaodZwQkpmu1BKyWtjhRq3qN/tKgF13GwqvnWhVP39P6O6S6ZjvrUL0mR1oWltLhf2IVAu3VvOckZP1z2u1eGb5+FIluWH0fKN6msN2hZvmH/E5nPxB23rBC1ezwvgh9TGRwkajjF2Fio1hqSLIwDNax/xbO3/Qf9SLAj3H7Av47a41uH7CM09/Cnjjk5h1Xwthfq1hNaoBikognBExvp2eaE0W8WtreQSb+YLs2mFmwbpenKraangiXCR29C/dfjAn3UQkOVBQHNNGgP/G7fbvhoFefKVIXA7zZ14ibo7pvDDG3r18hW9eQ/HfdgUzssT8rHApTZxms8cMfh4Ql9GE9+jACaE/GMOxoJVdmma5BaicY8GhVl0ucVkTrtEWG0mSksGU+ekTu1BC0q6fDdpqrj1eRMzOay4eXmuErDEqj1eVw20dTMhEL9T1ipdnA5eUfOKu0wpAzKvNVxHwxn3WFsRg85pglajMwxwyf2cxxxbDOYk2tr8GGAct4ISdUILMmstIa/fP5/TPsBPQci0OWhOBp3ZBoozuls8TigYT7DDGCbR1pAmXY36A3+7opJrIwI1s01e9J7evHpJr95+oHAMU/fGEksPC6YE/W71fOUxPzfHk2BCk7AYERBcugOXghaO/SnGKyuoWENWpmR+vC81v2mOXWTMs8Z9oe3CxL4MSa0Y1/n9lMIJtIlz8205v5qfFTxxY11r8In9Ed4ZzP5wgxjDzSXpY/PQaoNyRxC4EgRurZDG6Go6fIB499+cyey8qNehggqhgVAsETtGRxWC0qZ6WchhM9v6xP5mMQ3x4a3DPyrn2TApWtB8DaHd4ULYQTbCTsawQsxMPfgYKUkArH7kwUXLgpDGAlIEE2feboqdg0PQXkkGl8bDdVQaMMUG0LyUGHRgnlnHT1P42wyKL0XijcgH/thfeyYdVe3qACEZlwJJQpNRQOaSf816TKvv6vAdjEejwoyZkj/g3bCMuk6WgZ162ImrEsyUkEWLVKc6fMS84iuSTMDk1wFVZi//xQfC1j0ajgu04jH1sMN2SztzXgt26JkB3m+UXyYT7OYEHcW3fbnFhet5XANo0KA8UIk/amtNqxtrk4eVCGgG3Baa8D3Ut3ovs3pimgwmcMnC7tBN86QOHzbs/Mxd3TBme9ixGc5RA0Fx2Iqo3crnpWnDd3NYzBuek+WO4jUdqJ5jFxvzBtoOsSBQIx67KIeTeUJnBGjw/63BB1cNSd5t2e7TZtZazRPPrGsJn9TvU2cczYiZMyWvgUhuCQI7ReB2mjsvEdJ6o0vuqAWnbJwbxpou/Kuds34Dzt4J1OrmuANH49ZKaRNHNET9EGaoWc1md0itZhvCGj6oEI+0xpCGENzWpToLUdhyb7OQVXfhc2SobK1eA+YUOBzX8QmA8zn2FBBbRdYmHP1ht3FcBW9JnwypOZeJj3no+HA2tnsQ5GrYLj8qnPkXa6pBKNOhSbB7LOzp8PfAMR74bG2kY0O7ZHVpiCMEXBxnAZUUhE9oE4uNCWv7sMXNmt2nITY1NM0xBBXQqfMxKjlejV4PZlAIliM4dZtw0AfeOVw1crodGuKZNcA44KMN9GiL1nRyBzgUGX5HZh9+cTCnWgMcVbKNhFXWjsgV/b5HLnYxDKClSKFr7RabWqpepSq6jCs+7Hwsg6OxoG5Rb2RuQb9kuXQEeOMABBSe08rX08a4gsM/p8p5UrOhLcNRPhhHYxRKDZhY2aSqJKE8GW1sJoBflzuyMIb70Lq3sKkJfox1uFBg7KwXlVB2wxy7yJiv4Qig3qSF40KGUHrjqCHMmcm8SaAkGA77Q7eFY5Js5USngf5efdafdb+qeWJFXVX4uD34Z4JPTfAWkIf5mb6uQbknCFwBAt/w1tBdtnNyckI//fQTYasq3bt3b5dNSd3rEJjAvALBLwITv9bE54XZOOctWON/da2EXbTxFB9Ck/xuSH2RZy4K4s7K/fDDD/Tzzz/Tt99+u7M2vtiK2cSeQhNWrPmiPnwQa2OcLXiGFPTFdlgIFwR2g8Avv/xC3333HeEIDqLfO6uN/PmPRH99z2dwrD7b8s6tNnduiZFkEwQEAUHgq0GAjxOyCtMnq9t8L0G80EJi+2p6KR0RBL4OBMTc+XW8R+mFICAICAJbIcBHTHTHLk7T70GbBod7k03857Gxb9WMZBIEBIFLQECEtEsA8UZX4YwoWqOFvXKaNRdB5a+81R02WKMmzq+b983bYWNStSBweQjg8GsXm47cy6tRahIEBIEdISDmzh0BK9UKAoKAICAICAKCgCDwOQiIkPY56ElZQUAQEAQEAUFAEBAEdoSACGk7AlaqFQQEAUFAEBAEBAFB4HMQECHtc9CTsoKAICAICAKCgCAgCOwIARHStgEWBz/qCL1i4JBJE3+GYeKAQ3/tqfrbVEd8dlnL2yrr9WTiM8AMap/D0d9r6dScXBK1Nx6fS+qnVCMILCAwpgYOm76saSTgCgKCwJePgAhp275Dq0d+EFCAvxBhQ+qIebkQemXbeiSfICAICAJfIgKIC9tzB9XRCb7EPgnNgsANR0CEtIu8IA3hc8yI/DKI8UXqkDKCgCAgCHxJCCBGsBdEa+P3fkndEFoFgS8JgdsrpCEYehsxLTmQtG63aHQugSuhBDEqNY6MjpQGA8TBMxDLEaZQxLbs4ATvPOE07z5iwNmIm4lAjw0Ez16JpxgN8AwmDi6CoN0dByZVy6EGVqw9mBxbPtcUUR91dEY9BGm2EVqJG0WwbxexIhF82cI9dxTmdacjaiDu3dRS6bfJaIzVs3TcII6z2W81yQGtZqNPQUkQ2mY8TNTlILbftAtFT6b/qHYd1Sdroa+Mw5DaHLAeJmGn403NwWk0ohbijyoc7Aa1pwGdz4kP6nGZbvTZtNuI+bdMnFwLAjcMATVfECAdh8ZqBnhDHgX9nETm8ySf6xY57SI4Os9tZzhzuwg6ZNRM6oVl9SnidJrUVjxkMVXyrAni87reNDOHjLIHmGhBnxqIu+mHmIOIAdrxmHFU0AXu00H82s6gg5iyiEuL3625IO6x1wEfY76FP9RZ8pp01CALgdTbTfAXPGsgJm2AuMiOw3U0qM/MKp2QqzcRo7QkMSXPxblvkxXOutxluRYEvkwEOHbnLtOPP/7IQasyxO7cZTPnrPt99vRwT9E1/Ts4yd6eVlTz9nG2f/9F9kk9Ps0+vD7JDg8eZa/VjffZk8PD7Mm7vPDp25Ps4PAp7vKjJ9nh4ZNMPTp9m50cHmfPP+L364fZ/uM3quyz48Ps8Ruu6DR7/Wg/O3qmSuLRs+z4zn528o4vPmTPjvay/Ycvsw8FjR+eHWWHJ29RiovmdT/7wL9fZg+4Tb7Piel58ErlO331INvbL+n+lL16uJ89eJn36u3JYXb/BROHhLaPpm3nt8r/fnx+nB08ep23+wH5ir6+eXw3u/PgZaZqAD2PD44yRQ/uvLh/lJ2U4H58kd1nGpigc+GTZW8ez+hlOg6fFlgtkihXgkD2/fffZ7/++us1I3GK6be/yGf2DrNyiq8S9yp7cBdzY/kB5vDh0bOcp4ALvXp4kM/b09fZw/2H+VzimfbiONujvey4nMcZ5uFhMdcW6tzAs14/wvx+M83NfOZYMS2kT5i7S7xlLV3gPk8O7szxsqfZ0UHBk0Dzo4P7WUnihxf3p/zk9OUcfzp9kz3e38sOwCcVd3zzGHzsJX6fKr71SDEQTsz7HmavysvirvwjCFwFAizXKBni3t9l9Pf/uPr3uz+o55+TbqcmLZrQ+E+/LUrV/zaGNm2DoO13oDmCFseAtgke8ma3S7YKd6djdexR18y1ajWsGo0kUqvbNAgpgbZJPapZ1O53aBZ9JaWw36KhAc2WqiiCv1udmtCeqWQ0qVnUWdygZgcbDlQzebw9u2mRukTdTSsizz97NalZ0KIpujWybJ3iiNVRMN0GGjWcetFUk1oLbee3+b+BH2L1buftYkXb7zmoiVMNq2KHVA1YzTtGQqpq+LFMImy0sHJ8qG5TQw8pxLPz4QOKoblkermX9eaQRk1dtSxJELiZCIQ0Gv9lkbTf/kSjc2rTIt+nGjTUOWfQoOk2KeDwHZj3DTOkieJbKe5FZN03KcQzxQlCaKJ0h8qpNyOkmmedB8dKurgSRDWY8TJovsFjmB2AOPLBNxoFq9Ed8LTAm/q5abab8yfwEMuokw1rAfOXmgFtZMR11MhumOSPC1sB+ujXoXkr2Mt56Je8gsCXgMDtDAul6VS/g9fzn3OvaE8n8ITqxBsHYApQAkniwzToULse0ADcQdNi8gYDGvsBBJOQfLJUPUmSUE3LRRi+1lFmmmA+cH2TBn4h8ICtJhBxzCkNGuqdZUdFiLNXXicUJzDTztGr6Zpq76ykzSoB5ys5W16fOa2/Tlp9HddLKUUTGkw3eaqTVQh2/J3QpgTjOf6vPhTcLzbjou95sZSiMKEGHp4PHyKzA9MwTCEmzDyDQZsgY0oSBG4wAtgRru8R/WV+QXiH9PmJuwX1SZzAVDpjBjWewwkvVnihpdHAizA5EvJCDvfk0ACLyJAc0n2PajCHzrORvLlqnrUFOdMsm+jiReCMf+X8QvED9CWtg7+UtSCThr6UnGvGWzgDnpWsZo4daXCZ0HtjGFUtqnsTqjVGa/p4np5IXkHg5iJwOzVpGvwb+g/o7vS93KHjHoSmTULa/DvUoBVzNQhtLJoE1IMA0ktNcrsIVDyChm1evlkn63BdYE61NGaeNUu5VFNcQyBauJ4nYBe/59s7o+GqPlXcr9Wb+JBMaDLhP6yaIcj2jKIPFWXW4gMtXM8LaOgE1IbPivJRkSQI3FgE6uT2e3TMC8Ii7T9gX9OqQX/+jui2RYmH44Bin3wNGiVo7u2aT16Uku+lEOLWMbUzeNb5ybjaEtCcNTQPQikEU/iiOWv7eLUkSWuCwK4QuJ1CGtDU3TGFH97Qq5ev6M17mArbZm7COwfSSkSA+n6MUMX9NjQ8WCFrtZmwU4OmKp1qt2DenGD1VwplOkydnYQ6rVFuBsBasK5BoxWXBECAm/5eJgp5a4U5sXiURFhtl0vXBekO9CwXX7nO65sp4piOdaVqSvmWJOUz4DbyC/pXKi1uoAwLo9PHYKz9vnL6Px8+CQVjaAhSaA9aQ5p0NeoP/KpG5b4gcCMQqJltCEzv6c2rl/Tq7QcKxzDvnZMy1n7Pa8lTXtnh3EYl6kEosyKY/KA1S7HpB/YAatiYYzCRTiIDbgdrGtvAs/JKZ3N/00JxI11rmlW30JcaGNt0bQqmk6Avq9q+qgpUJeRA0B1PRjSOsdFpXR83FZdngsAXhMCtFdL4HWk6/KN4N+KcKWGrdwdz52CUkM3cgRkMNEOQkZAgiA0HhAWsEow0+LDVoEEK1COYRrFzad5tTHeH1MX/3GGEDDp2O8U0GuX+JIkPn6vKHacaeLNGHvwyFDuF39fIh/8GO5/UYMpFi76SirCaHnlnCFGcD23DXDIufWWwi3LEhK6R0wzsAA3G+CBwMeyQ7cDsMK8M5NsLCb4ltu7RcFzkisfUwy5X5srnwwc6S5g4BwrMXMDjnbmSBIEbj4BmwI8KfliWfu6FIPdNh9Y4GRc7OjHbJuAR5lQywW5PI8CCJcI9U0FhYid1NOxRAI2TtQ6cDTyL6gbV+CxILoedqePS94uvsUKrQWor2cJmutY1jHvY4WqBb5Q7s2MIWhH78VZkr7qtOw4lvS6F2PWZ97oqp9wXBL5sBG6nT9pF3pkPM6bZB+Ni37EamdDmDJVds0k9bIl3cbxEX4efmNMgB+YGxchwFMbQbcF/zYb6CR4kbZg66vONY6PAoEsjy6WB5VGrO6RmC75W4KyaCQd/1L9GTlIVGGi/3WrhSAvwTmTS20PCof9IFnW6OP0flYx0HQwb29eVwLY5WTDVmq5L9hh+ICy8YqW6TkbkIzw62ETh2D1oCWvYJj9SDLa6CZ1c9NF3QUefTbzYYDAYAiMUOic+bq9F4xZWzhreQWpRbyTsefNbladfHALJGJt2zKlmSYemfdzq0rDRwvEztprsNbMHv7OSkWA+YTI123Xwkby3NfbZajYp7Q/XC4VwP6jkWYYL9w3etDShug5NHNrkqaoSb1QwutSEEz8fj9F3quiq5gZwksMiDbyuYdEAFac4c3IwcFQbVbyubH7hXwh7jtanGJsIJAkCXzMC3/DW0F128OTkhH766SfCVlW6d+/eLpv64utO4whMSy+cZRP4XtkU4lC0nvChL/7dSgeuFoEffviBfv75Z/r222+vtmFp7YoQwFls2MxlQRPXEIX6FWEuzSwj8Msvv9B3331HOIKD6PfO8mOiP/+R6K/v+QyO1Wdb3rnV5s4tMbqybH7XoeaoWIXChOlF0ITpV9a8NCQICAKCwJeBAPzqPN4oIQLal/G+hMoLIyDmzgtDd/kF7W4PZ3/BLDjgDQg1MmCClFXi5eMsNQoCgsCXjUCEneLrjxf5svsl1AsCywiIkLaMyHVe1xs4qqJxnRRI24KAICAI3HgE9DaO4LjxVAqBgsDnIyDmzs/HUGoQBAQBQUAQEAQEAUHg0hEQIe3SIZUKBQFBQBAQBAQBQUAQ+HwEREj7fAylBkFAEBAEBAFBQBAQBC4dARHStoHUa+H8s9HmQ1u3qUfyCAKCgCAgCAgCgoAgsCUCIqRtCdSNzhYOyO35N5pEIU4QEAQEAUFAEBAEzoeACGnnw+tm5o4QXDk613ndN7MfQpUgIAgIAoKAICAITBG4vUIa4se1EYeS4z/qdmtDnMwNowWx7YaujZibFv5sckehCm0S9nDdC6cF4wFCq9Rbc1vGERbFdGmyIldx7E8XcTRN1GeSYSEMUnG2bdBB4ORRWYCjEZjUQRMR8tudCcJoIpyUU+SvoIviAUJbtWjQc6mB+H4mwlENp2Qm5PebRV8QDxChrqKiB55rkdvvURPx8izLoc4koHEb4a8Qesp0OuQlyOi3ybAHc+GhcCI44uohxKkkQUAQEAQEAUFAELgAArdUSAuph4DH/8e//Bv952//SX/5l/+T/rdGZyH4+TZYRgOX+lqPPB+aLK9HGmJLIs4xhBUERPb8wocNQc69gH77D+QpBaLQp0CHwDMNile0FiP+Zo+o5yNAOv7GzYDa3c1mTA7S7vUc0ht98iYDatQhuFXQpVoJPQohTI0nPk1aMXV7kzxmHmKTthD8b8h98SfUijvUnkpYEU08nfo4QNIf2jRpQPhCHRP0cYgYfx0OEG81yY5xCngplEU4xwgBnuVE8G1GkuQRBAQBQUAQEARWEbidQlo0ofGffltE49/G0KatAlR9B5onSCR208oDECP4cNOKILBB22VCAIOQwj+JAvICg+4fx3iWSzCx71ENgtxKRJO6S2Mf2q5CeDOgUaOkUKVVE7L0ZANdnBPCYbOQDuvQ1mlxrIS0CMJZrdlQwdIR3p2cpkkBNGZlsvAM8h8kUGj4NPTVUVfQxhmUhEyjiXpDGnuq0xR7fCK4k5dRdyQJAoKAICAICAKCwHkQuJ1CGoKY1+8swbSnk5HLHVvil1CcwFQ6V0bTNUoSCGIQ2BwDwlmIquAvFmhN6rYsKLF8CEQpBRDuLFtf006NtDSgYa9NLQhFDZgxz28t3EAXt6gh5NS0ZS0XMHGdxAlp2uxJrY7fEBBzkQt0aXNqP4Ssml6q25yrBmUaBLsx9zEhb1zVxzXdlluCgCAgCAgCgoAgsILALRXSGtTvP6C7Uzju0HGvT+65hLQVLOduaPAP02A2jCkJYPa04bdmwyQJDVoIIWwSmgSXstUUDeHn1aHQcKnTH0JY+7I0UTWYPI1gTAGCw08iixrr+rjaa7kjCAgCgoAgIAgIAmsQuLWxO3V3DN8s+E3BBKmZNtnGTIu0Bqc1tzSq12BaZEufnj9OItRVOGHVYc5MuzB51iOyWFqp18jRsHnAryt/tN6aGhNvRKHTg89XId2ErHcrE6usco3VmqJztzbTVVVWg+ZMaQELPVsKzRpr3eb0Z1VFZ/c1m5p6j8ZDbIAwmzCAShIEBAFBQBAQBASBiyJwOzVpBVqablMDZsXzC2hcgQbXMw1mvSAXpKA9GkEAs8vdAEaDzHBIfV8nbKREMqAlS2gMR/0abqwTfpSJMQyLHZIxjfoTiiGXsWhWhy02xAYEJbRhZ+ood3greoI8SSnQnUHXtMTiDx07VJNxuaMzockoQP+MitxVt6FBbNRo2J1gI8P6PlaVlPuCgCAgCAgCgoAgsIjArdWknXcgpH4Hx1eU+i841g886rWG1G61CNZMqkF60tvYnamXNcOkafjUjPs0LCQyE8JZ2Pep3V9vV605XeriSA3HGsHXrU6WA0d++HYpIa3RxREfLZhN66RDE2fi36nuz4Kg2Xbh0w8N3WRIzSq6WOtXlawuDRstcnGUCHemhr4OmuvprKqC79ftJunpmBrl7odNmeWZILAjBP71X/91RzVLtYKAICAIXB0C32RIu2zu5OSEfvrpJ/r3f/93unfv3i6bkrpvAgIRdqe6RCOE0jq/iHcTOiA0CAKCgCAgCAgCZyPwyy+/0HfffUd07++Ifu+sFvjzH4n++p4+R8y61ebOVUTlzuciEHtjSh1bBLTPBVLKCwKCgCAgCNx6BERIu/VD4DIB4KM3YuxsPa8v22XSIHUJAoKAICAICAJfBwLik/Z1vMcb0guNmthI0bwh1AgZgoAgIAgIAoLAl4yAaNK+5LcntAsCgoAgIAgIAoLAV4uACGlf7auVjgkCgoAgIAgIAoLAl4yACGlf8tsT2gUBQUAQEAQEAUHgq0VAhLSv9tVKxwQBQUAQEAQEAUHgS0ZAhLQv+e0J7YKAICAICAKCgCDw1SKw892d33zzjQLvH/7hH+hv//Zvv1ogpWOCgCAgCAgCgoAgcHsQ+K//+q+8s4Wcs4ue71xIK0/affv27S7olzoFAUFAEBAEBAFBQBC4PgR2GLhp52GhtkFNadt+9wei/+W/bZNd8ggCgoAgIAgIAoKAIHCzEZCwUDf7/Qh1goAgIAgIAoKAICAIXBQB2ThwUeSknCAgCAgCgoAgIAgIAjtEQIS0HYIrVQsCgoAgIAgIAoKAIHBRBERIuyhyUk4QEAQEAUFAEBAEBIEdIrDz3Z1b0/7//b9ESbR1dskoCAgCgoAgIAgIAoLAjUWA5ZrPTDdnd+dndkSKCwKCgCAgCAgCgoAgcNMQKI8iuwhdN0KT9k//9E8XoV3KCAKCgCAgCAgCgoAg8NUicCM0aV8tutIxQUAQEAQEYsBhUwAAIABJREFUAUFAEBAELoiAbBy4IHBSTBAQBAQBQUAQEAQEgV0iIELaLtGVugUBQUAQEAQEAUFAELggAiKkXRA4KSYICAKCgCAgCAgCgsAuERAhbZfoSt2CgCAgCAgCgoAgIAhcEAER0i4InBQTBAQBQUAQEAQEAUFglwiIkLZLdKVuQUAQEAQEAUFAEBAELoiACGkXBE6KCQKCgCAgCAgCgoAgsEsErl5Iiwdk176hmqaRpv4cGv4/bTKMNvm77GlZd+JTv2FSnds2XRrF1Y1GfYvqzQmlRZZ4YJPWHKvr5WcrtUR9sswehSsPLudG2DPJ6keLlXGb1u7avDDlCosOBReu4JoLrsU1oZGjkTNMzk/cpY+Nz6Dl/NRLCUFAEBAEBIErQuDKhbQ08Cg0n1GYJJSovwm5/2ufwhAf8ivotN9tUl+DIBMH1K2NaOiVIthy4ykFXkiGbVJNPSqurfxab/sUj5zi2XJZ5EY/Y9MkffXRJdxJQFtCprlYe+qjTcMi4xJauMwqFBaGfUPoAnYTn6JzdHA9rho1JwlNXG2rmtLIIy/Ix9rlj43z0bIVwZJJEBAEBAFB4NoRuHIhLfQCqpnWjoSXs/BMSXN6NOw5pNUSipM6GXougq2WDMkL62Ra9eJRhOsarvXVrGvusICnFwLdmsefeSsALSZBflxIgZ8LlZ9Z+aUXV1hMhd1Lr/58FQY9cnvn09l+Pq4pTdouDZN8rO12bJwPDsktCAgCgoAgcHMRuGIhDVqMICbTntf1xDTAB7xT2MLCYZNMwyCz0SRHt2nA5sgA5lB7QMoyyeZSIzedscnRaLSoaRrkjqGliD3qNWyyIRzpRoMGK7bGGhlOk2zIXcm4RyNqULNKfRf7FMQG2ipeHsykQQTBiK9TlNRd8vhR4lEHNw3TJrcB4bPFd2P0M6Va0KUGhDrNcAlKl6UEzdygCTOgqfrr9AsAPJfqZpPaLjRP9TrZ/aITaUhD14L2DDQ0uuRp+L2gxInQJoRIdTOlcOgCB+ThuiGUoDXq4HeOc0LjpkbfaKCLqUp9akPTNW899Vy0rcBHb4Y21b6ZL2tQY4QOAe8O+mxZ3H+YjiPOzaY3nZyWSxbewTBmLEjRFY9dvPse+XhVSdCnpm2j/yhb3Fstq5qfptjrAU+8c71OVof7xCmicdsBDSZM5ha1C6CDDurF2HAdjIU6xhcaTf0esBtQGI3IbQ4p5HEFrFsOxlvbq+jPPK5zxHgt0hsj9BbDc01bs5wRjfDeWv+ckt91qONxfevHxv/P3vvANnad94KfHxyLASYQ7ToY+k+tG2ACMa4fRGMdiwHGFdcvXTG7qUV3HYhxu29YtGuxgZ/FuK8ZGvC63GwehklfLcY1MmzeWwzTdiMGcSt2GmAYNI0YxIBo14CuUcfDIAPMtdexGMQZXcOD1R3HwN3fd869JEWR+q8ZjfQde2bIy/PnO79zzne/833fOV9vTDraxLyqpNH/cITiqQQw9UzbHbQ4wDXBmEaTwDdESV4Xq5I/N4B92KBopqr6wNibBawNXjvAK+1P2J5j3L/Pdp3HFXMCNEQYa26+WaIY5msiFacI5nQ0laUM5k04ZFCiZOk1HUyQTyq7FYRTtZabQVcH5KsgIAgIAocHAURnv4pp3j1xlNyBgUF3cJD/jLtnLp5zJ4cm3XMrIOP8KXdkZNpdwOeVcyfco8PT7iIeL50ec4en+RM/n3SHJufcFfw3NzHoDk+dc5fUL8vu3OSwOzGrv82fOOqOndaf1c+daWnWnRgacqfml7t/aX3ndgZpwKNzEDSTSyOn3POcY2HaHR4/gxZ1m+NnLuLhRXdmdMAdn0WdK+jT0RH35CJ3atGdHh5xT6mCHWn5nHtyckb11Z1HX1V9qGVm1D06Mav6tDI74Q6qvurnw1PzKs/C9JB79MR8Z23IPNfGcfEkcDzpquYvzrijR0+450D5qRHQxHTg2djYGJ4Dd3xdnpt0h0+cU+34aX5qyMNv0T05OuaOjQy70wv4dek0PmOMQMnshI/3insOeI/OXESGeXfq6JA7OXtR18dYDE24Z85Nu6NjJ90F7sDKgjs1POqq7MDn5PAwsOLPXWX5kZ+W0b/hCVcNL+M/OqPG4iLPDQ+X5dlxd3B8Vo3LmfFBd2xGg85zYfyMHuv5E0OYI7qnPK8GRz2a+vWnE9cOcs6fGnVH1aD2b6uVXWGG8eAH/eZGX0zajS6dGXeHMR+4J4vTw5gDesxatHh1nDqP/i3PuuMDPsarCG/PDdDFc2OeyZqfcofHPEwxB4fVYPcb4z59XsH4DY+5py9yezzvMWe406j7KPrPc335DGN+GqtFz+mhKbQDjCeYF3hzY3rEr4PrkSQICAKCwOFF4MarKo42oI2iKao60Ib5DZtZykNjFoElyCyVKJCqUhSfofeAWZT9mByqQfsQSWqVFptLjVgBvmDa5JcpxgmKMRSoUanahJYkBWduh5rQVoQTq1RNukVojbLxLNnZKpVjPX736OJ2glM1sopa1caO+rGG9vey6myyTVEQ/nSlOrQ3JQOlGqAUmjfWZDXgdwdNW5475TSp6eA5Z+lMwQgl4iaVYQbL16sUjGUoqPrahBYqofrUsJrQ9ITR1waV4aCegEZLUQztRKTb1snYejjWSxWFIzevMuOQRAi1BPBdtZEvUzibJSddJseBpgcqtBRwVNm9FODMSKxxrMdylGqkcLAD+rgCxihToahdoZwZpWyZKUUb3EwI1FkmmaE0lZKGrg9YmKDfTAcob5Ypyh2olaiCvhnQrLAep9EMUc7A8+6yHi2KjmqJ6hHUy82F4MNY56cWFUpNSvCBDs4DGgKgIYgW6xirVJnnjE1WMwjtE+doUN2EhjDHfUNf6hbFMQ8UTehPqVd/OnDlFnWC1qkOjXCG66/1aaudm33aLGhTOXffuVHvg0mrGoxTEfQWEqqv6KrnL9mmxYEWtmqkyQyjfwwGNGLd1vlGBRrEZgCaLMxhu0ENjFWI50SpivHIaxrhb9ngdu1Sb0z64OvUilQNp7GeuTDPiSAZPCWwlkLQXPK6NhvADVpAzlLD5zCrpvE8FMSKR5FmBVriaM6rg+uRJAgIAoLA4UXg313NrttwILciMMF1NNrUtjC8KNgUakMoMdSvjZZPF/uGwSyjJA6YzupsOsObGo7/Jo4atFzG8DJt4AVVgcBTrUIwaMLsluAyncmC8JCkWrxM1TQLP/1St1mWX4QQFJVgpF/u6uXCAghehPxOhP2O6k28FPHYRp8cmOUMrp7zQHjiF1Q74aWYjlHGjFAmnyfIM54JmAVPv6/ey1dXqOpWplYIJnUz6Jk12zU26woY4KhNrf6hAhvPLSXsaiENpx2o0IA5FaawQIBNgAUIPhkCHKsSC2ksHJYKDqUyUfjwoQMQSgs1mA6TwB9jaRoQWlW/MEYs/EBA1U7xWpjlXxQWsTTFA9w3Fi1YFoOAkILQVcNY1UAfTMYpvMy7y3YSZOJFv8avDQJ3rRFWmPO41DFn1BixsBcE5izNYJaoPDx0PEY2zMsGP2eaDdDsDUyf/rRx5TJ+4rJenf3a6swNod6nvd/c6IdJqxoHtFtem2qMCX3COKh+6OeqDswB3e0aDuh0H9bguYtNQBG4VYE95rSNTVIYc6ptKu8gvA8m/fBVGxsArdq36hDqsT4NHhcIZgp03/TNdHObQb2pQQkjiENEWNOFIlEmB/N6BxnyURAQBASBw4rAVRXSlOAFIU0xcZXw0sAzPkGJNyhOeupnrBUrViCwsRoAu33TCsFPBr/Bl6jkvZD4ha6FD10TF8axAPWlWYJ2DT5slveT/semWiZB+WCRKvlo6yVg4yVrqXY7E17seIFHlX8XJy2EaH+vjpc7tBXQ96m6rHKRaqzJwmfVJyUVsCxpesKT+uolkyo1nMiDL5XBGpyaJ5itEjw7BAEb2jiPYscsUglChxbY/Pp8wZHbbEJzhH/UWw5CadGE/xJrybSQVleaMGh1IHQFgXm5UKdYNtkxJrrOQDAAmaxAFVydkQwhL+qzisA0mcUVKhg5Ppnr0aTywb+NlZ3t8dT1qO/xFGXSQSrnKyiDsqwyYaEP36opA1o9YITUXVbXoP/mk8A6wbcOgogq0kEDVDBUrME3MQ69EJ9yRR6DC7AmLOQJySx0wG9NjQwENtPxBbZ+/enE1Wue/4EA4guBfdtqZW9SHf5wYSVQ9Z8b/TBpVQPMWDnGyTFLVISAr4TTTlqaXgbgWilCO+sLTK1K9BpTUwMCbgY+aekal4G20YfXgpYb2jB+3G+M+/W5PUbArVgiO5GCmMZriTXMPGm0wKz3OvjsbWqgR6NQiH3ltBaN9wC912WrI/JBEBAEBIHDgcDVs/Syz9ZR94RyPvMT+0m1/WbOnx53h4ZG3PET4+7IIPyPOCv8bE6ODMKnZcKdwvMhzy9s8eTIap8zzjc65A7BD2Z49IQ7q92R2k1dhL8bwR9O+cLhz9FR99QifGiGfH+oTrKQ1/PVUU+VX5fnN8f+RZ6vnAu/sqmRIfjyTLiTY/hXOVahT6PtPp1jH6i5zj5zhUvu7OSQOzg05k5OT7qjg5MuZ1mBb5j2t/PahB+P7sZ5d2bsKHz0xtwTk2Pu0bHTnh+eT3Mnjivu4swYcByG7xGwPKX92JR/0fiAO+D7RikfNfjcwReul+ce+8YNDLR9gxbgo0aMgXbtQhfm3BPo+xDaGGnhveSeHmNfvE66PH889pEa9D4vzrjjw8ALYzUycVr7zoGK1WX9OvS/S+em3BH4EY6MjrsnTi96/nPn3TPwixuCz98I/Jimz+meMK3jZ/TnZfj1DcFnTaUF+OodPeqOnJhzl+AHOATfv9bI9OzP6vnpU6TGyfMH69tWi/wl0AgfQoz1qYV15kZPTFqV4MNFzBn0Ff2cnIDfojcHOmlZQf9GMe6jE5PAF2ttvrO8/nxxFv6HyMN/2nNjBdCMqrU3Ojbpnpr3BrknJoCxD77L89w+xhVzYtQfV147w54/HvucwpfwIpPS8uvUdJ07Ab9P9ktTw9ZnXa7tjjwRBAQBQeBAI3AD927fiaONPPxWAjBdZrQ2ZN8R2E2Qg9OSYZzig+9YXOkpJAkCe4YAa4pjODncKET7tFGntJGjqIk7CIN9suyzx7V0iNKBCnwe21rufUaikCMICAKCwFVH4KqaO9frnVOBQ3S0AAMdTCWlMlEMpsD1Clzz3xo4GMCXiIIQmNpK9Qiu8xAB7ZoPywElgIUYvkCZTdiVkoWrRJTRtp1UJI848c0oDg5m1IwEISDC9ZFg8i9Wo5TLioB2fQyYUCkICAJXC4Gre7pznV4FIgmKNeHvFC6pe8WKJXa42c8JTue4D6yQwn1QOD0ZxYnJxPXyUtzPsAptPRGIxOFXmE5SGL6MoXgBJ5O7NgQh3EOGU5EZOKrlAvDZK+O+vZ417b+HjSLue0sW4fu4/2gTigQBQUAQuJYI7E9z57VERNoWBAQBQUAQEAQEAUFgHyCwb8yd+wALIUEQEAQEAUFAEBAEBIF9g4AIaftmKIQQQUAQEAQEAUFAEBAE2giIkCazQRAQBAQBQUAQEAQEgX2IgAhp+3BQhCRBQBAQBAQBQUAQEARESJM5IAgIAoKAICAICAKCwD5EQIS0fTgoQpIgIAgIAoKAICAICAIipMkcEAQEAUFAEBAEBAFBYB8iIELaPhwUIUkQEAQEAUFAEBAEBAER0mQOCAKCgCAgCAgCgoAgsA8RECFtHw6KkCQICAKCgCAgCAgCgoAIaTIHBAFBQBAQBAQBQUAQ2IcIiJC2DwdFSBIEBAFBQBAQBAQBQUCENJkDgoAgIAgIAoKAICAI7EMEREjbh4MiJAkCgoAgIAgIAoKAICBCmswBQUAQEAQEAUFAEBAE9iECIqTtw0ERkgQBQUAQEAQEAUFAEBAhTeaAICAICAKCgCAgCAgC+xABEdL24aAISYKAICAICAKCgCAgCIiQJnNAEBAEBAFBQBAQBASBfYiACGn7cFCEJEFAEBAEBAFBQBAQBERIkzkgCAgCgoAgIAgIAoLAPkTgxr2maXp6mp577rm9bkbqFwQEAUFAEBAEBAFB4Koj8MQTT9DXv/71PWl3z4W0G264QRH+qU99im666aY96YRUKggIAoKAICAICAKCwNVE4P3336eFhQXy5Zy9aHvPhTTXdRXdL7zwAt1+++170QepUxAQBAQBQUAQEAQEgauKwNtvv0133HEH+XLOXjQuPml7garUKQgIAoKAICAICAKCwA4RECFthwBKcUFAEBAEBAFBQBAQBPYCARHS9gJVqVMQEAQEAUFAEBAEBIEdIiBC2g4BlOKCgCAgCAgCgoAgIAjsBQIipO0FqlKnICAICAKCgCAgCAgCO0RAhLQdAijFBQFBQBAQBAQBQUAQ2AsEREjbC1SlTkFAEBAEBAFBQBAQBHaIgAhpOwRQigsCgoAgIAgIAoKAILAXCIiQtheoSp2CgCAgCAgCgoAgIAjsEAER0nYIoBQXBAQBQUAQEAQEAUFgLxAQIW0vUJU6BQFBQBAQBAQBQUAQ2CECIqTtEEApLggIAoKAICAICAKCwF4gIELaXqAqdQoCgoAgIAgIAoKAILBDBERI2yGAUlwQEAQEAUFAEBAEBIG9QODGvahU6hQEBAFBQBAQBAQBQeAwIPD3f//39Oqrr/bsaq1W6/l8sw9FSNssUpJPEBAEBAFBQBAQBASBLgR+/vOfE//ZiyRC2l6gKnUKAoKAICAICAKCwOFA4Pb7iY7FV/f19e8SvXN+x/0Xn7QdQygVCAKCgCAgCAgCgoAgsPsIHFJN2q/J+nGNMt9+jeoXLpN90xEyjhmUeSxO6d/68O6jvKUaV6g0nac/NIkG4ymyn/rYlkrvLPMSZSaL9PUm0chjj5P5+x/duLqf/TOF//hF+mlnzpsCNHTsGKWBZ/bej+CXbdTr1/feMll0MxlczZUlKua/RwWzSdZlg0r/8L9Rkp9LEgT2KQLOG/9G2edfpMrr71DzfaLALbdS7Phxyj/27yk8sEmilxYomqzSS3SEpr7xZ1T8LSL7X/6Wbv7yBaKb7qG5f/4cJTZZ1ZayLf4ThTKv0C+6C90URpuf35s20dZV6Vt3n7bxvfmP/41ue/atVSUHjgQpfPc9lH/ydyh+G37qMXabbcpZWib7tpsphALO0r9RJl+jygWb7Ls+TY3TnyJjsxVJvusagUOpSbN/XKHo0y/SP75ugwEFKEg2/dQ06U8yf0eFpet6PK8p8QO3BGnolgANvO/QG6+/Rk996QUq/WqbJF35JZX/8gwZv1ek3Ou6DvuHP6DMD9+in14KUOR4mMI3bbNuKSYIXA0ElhYp8YUX6OsvN+mNyzdS8AjRu80m/eMLL4D/LBL2QttOAWwsh0NYb6Ej4F97n9TaVu3pNgN73+R11MKNNMi43HIjXbls06svv0ifefrH1NhuD5bOU+bP/opC//F7VLuiK6mVvkd/bb6D99WtFD8euipjvl3ypdzuInAIhbT3qAwNmtod3hKhc//wZ9T8m8/SGL/w33+LSi8u7y7Ch6a2Gyn+5BfJmnuKml+L0BD3+32Lij/cJp6XUfasRW9A++AnB8Kf4lmhe6jw55+kyGY1EYdmDKSjPgLf+MY3aGVl5ZoCYp59kb5/mUm4laa/8Z+pOZelhUduVTS9ixd56Y3tkxd44GFqfAfr7W8/Q7HtV7PJkkco+Qza4vZUm79LXd43m6znoGa7k/LfYN6XpcVH9fjSBZNKP9tmf1G2+PI79G6r+K/JvvyB+jZwf4wqv/8xEdK2Ce31WOwQmjubVHtTD9Vg5B6KKXNZiKJ3HaHGJXy8xFz1A8r/0fP0FKwJY48mKPL6i1S9K06NP/042YsLlCmZVLvwDsykrNo+RtnHP02J2z5E1nf/G33seai/7zpOi3/7OxRBTfW/mqFPvWDj0510uvy/UxoqcOel71LoS6/Ru7fcRwtzv0tRqLKTT3+PKs0bKXx/lKIdggn95BwZX6jTG1z+H1D+NzrK04008cx/psp/gIn2vUWKfbZCP0JfTv3Nn1B2aIXMf/wBZasWmVCRUyhEkch9lH/8Xi3c/HiWAk836AoE1ZlHP6DyWYdSX/t01xz+NdX+skD/41lg0tGnrkxrvgYjx6DlMpWAZV9y1vyuH/yaGv9SpXSpQfWmQ4EjbHIOKyyToQbFk99DXzg59K0v/Tn96P4gWS8zjkjNOn3qd2xtcmGN2zd/QIWX36JG8wMK3nUnxeIxKnzuNxUjq/3lXyj6B45/mopHGpR/06DyYw7FlRnnTpp+PET1F16jxvtBiqcSlD3yCmW+CTP4ZWjr4nGq/KdPKHOD/ZN/xfM6VaF9tWHODUciwNIzaWiq5O99hMDs7Cx9/vOfpw9/+Fq5L2D98brjFDpGyd/6kPoYffAYDb3wDtazDR6ygnX64dYcpWP30am7m9jYNMk+cieln3yE8qO97flrTIK/+leK/R6vmQBNPHqc6EXMVajqwpj35ey92rR6ZZkq36xSXq0VJgt8j9fKhF4rWxu+X/bnkT/xeCRMvPaRWykC14fM4zHFI1X61c8oDxNw+fUmNbDOwsfupGTKd43oTYXzs3+l9LMwG8M9xcH6M7DOEyiTG70ZPW7TMvQgePGlV6j0+mUKHItQ8ZnfRbu6zk7e3YTpOHx3mHLMu4e2T9dqaj8EHnsnDX6bBazL1OT3CbSna9J7F6nwbBU8i03gN4I130rRhz4NnvUxCr00R0Hmy1zo/Qv0+f/pq/SNY/8f/RjvIk5XflimGy59lpa+/kkKLf2Mst/Eu6kTx0eBo5ozvcenEqrSJ775Dvg5+P79NhUwSZxb4Orz5Kcp8nKVMnhfWNhUJB97hEqfYXcXzJm/rlLuhxbejx/AXA/c0UZx4jbgLulqIHD4NGlXHLI9ISgEU4GeaL9J+f8bGrW5PyNz6jdX4f6jFyr0daiZ2Z+EfrUIYapK38L3APwO4qHL9BKY4cNfqpGJn42779QapOY71FCrbJnqeKnrBOHwwq/VRwsCHu+SBiHMRGiFKs9/j75zAVqiI2CaN12A/4pXhP+Br1xULXSb6p5w2QCT1busD6gOWlR6E4yX/w1BSIEay3mpSvFnX6Hvv/4BhR+8h8KX36Lvn63g2c8g9nSkS6/BZ+Y1eulNZ/VzzvKr1yhfZaE1QOOpqBI6N5Ns8wKZHsbBUJ+l/LMaJb78Cv3ozRspwb5rdxPMBHX6/NPAciBE6VRYYwlBdATC0tNjv00nIaiphBfY1GP3gZ5fU/2bZfr8Cw166TKErAfvBA4X6FvP/x2l/uW9VaReebFGf1hlU6nekeofm8D0TkqhXudSk77z7H+n1MshSh2/lQIwW7z0Al5oP0FOHvcvfQ/j7lD4kThlmNYXXwT9/wqGJkkQ6IUA+Iyn/SD4obGgr1Lr8wf4vWsDgzVs3X0f+AqbRS366jfNbZjMHMUnEqgk+P5levWH36Psy5rvmKW/o4fVWsFL+CGD6ALWyrO8VnamcVzNI/+NUl8Cj4QAEsCmMAke+SOsFZ9HMk8sPF2mp/DSt46EKX38CFmmdo0o9nM1uXKRMrz+IHgZD8apEL+V7Ncb9NWnK1TqKvPGhSaF4/dQ5KYP6Bevv0Lpb1/UuK/i3RFKHvtA8ZuHn/4B1T1evWW6/DFt/Qt+1OLNRyh0y5oMeLBC5XyZvsiC+F33UeGxeyh06S36zvNlSv8Y4wDhKf+gP1tupcnH/gP98f/6aZq8S9c1AMFz5hEDG9BfUuHLZfqqwtGgZCSgcQS23S47q8bHJwlzzURbCdT7izcb9BRcfUq33EOpYwG6Al74redrVAUu9b8q0cPfblDjpmPwo7yHjEuWmjOZRT2nevVQnu0uAodPSHv/A3I8ASJwZBOKxJsMOvW1NJmPG2SDmdRYZjkCc9tXHqbyMzEa4fHAJC+zapsFKmU2fYdMFqjeewuCGf696UYagHZOC1TQIGGHSRA+IvffCfHnHeyENLNmrV3xqTilvAXJVdMAhK5j/OEyWW+y4LFMNSX4cZ1YYBfA2PFvE4Ifm3AHodlj4aVWbajvA/d/mipPPUyVx47p/C9CS4TnncmIJ2jhb5JKy9eZzGqdvs9YhSKUO957R9/O/wFVn52h8ORfQEtoanMysEvdf/PqSv1vN0FT8HicZp5MUP5zn6Tso8foKP8Gn53Gex/FS+aY5xgLnB78FP3RZ/8HSt/vbUshpKU+93H83qTyi1pIHX30ESo/laTicRYKHaqCefnisW7yCE08maLFQqxD2AxR8tF7Kf0IhFiVCYLeI5+k1KMRLw+/aPGYGRr/i21xPA4sHovR1EP3UeruwFrBVjcmfx96BNp8hsATWluVjs8+H2pBhZdm9jP3UuZBz2R2GYeatoxjgGIPPUCp349STPlsgg4lLC51rBXwmf/0OSrGeT3xWrmwQTuX6VuZP6cbxrw/f9Tlb7WKR5qk9nXMI5/5DDRZMRpjMnweuQReiY0jr7V09mEqPPVZSjO/g2tEf1cT9Anam5nHP0vFxz+JA0le397nTRZX3k5DsESkP3Mca1M/8wXl1bz7d6mUPU6jnOXN16jEPHpbdPntvkW5L/wVGQ//V1hNND8auCtMyY93Utb+bByPoS/QRkF7lf5cjNKKvztkmhjt2z5G6QeDilcTLDWJR+6j//g/30eJkPeuuuseyjzwUQosXWjhmMp+nkr/5bOUYtkOLjvlbpedjvFpUQH6MhP3UvYhTyC8BQe9Poe599Cdum24ltj0XksbHEC7KdCaAz1TD8EfGPNK0tVBYBNSytUhZL+2cvTBGNTHWnppNC9rNfQt2CXxKsK/Bhjhq9ixKtX2xyFQgTl8B4ut0cROA35VdQg5w3ihGyzwQKCClQGMhSd4iOJ3s+CDHZUSGuFYDKd7TgbvwFrM5yMUvRvZ0ROsAAAgAElEQVRMGxqzBgQxgiZQCX53Ryj1/iv01xdgzrzyKQpC5a0FPwMvhA9Ajxb8gr5jsac1vMK0rlIy3YlTrfdSFGZUFgRb6fIFpQrnOscehQlWcY3105VLOIDBWXDCaQQCK5/u7Bb8WjUMHaP4BZhjX6hT7vnL9C6EZ5VYiF6/mY5fHY07aGStKNGHvH+hlYTZetULDkwmP/ExLYxxt1S6kQJKqG5/D/J3CNWrEsywsVteoe9gh/nUF/47mY+CWT3+u5s/nbe6NvkmCPRGQAlVEOi651/v3H2f+gLh6rmNTZ4/77ERqr6EzRDzLKQr4Gu8Vjw9dc96+eBAyKcPfKoldCJ3J4+0fB55k0ONl3nn+o63nm2yuL33be/AxI3YWJ6HSRZr2GuxibKEk9xr0sBtFI+8BdNenQpwoG9C6FTKLxZAuzL72AWwtvl3P3HdqozPu2+LUOkbd+p+s5D45jboatUOrR2sJ5wGoCmNgDfnn3ygj+XhwxS9Hy4WMFPmv12jFPi07guX3oLgA16rcYPLjdLYef/iYROYMm/z06p3WOup/qBx8j93z72P4CRyiAZwmv7dF1+gyDQ2EdigFn9row17VyPydUcIHEIhrf1i1jtMjZ/z3ntaWIK/Q6hjDgaOtNlRZ3791IcPpov3Wf3bIVBBW2YeeQsapSOUePA+Cr/OQhoEql8FCHMeMpqhzJJqYfqaPQgZvKPqZICcgzVu7OfwizffIQvCGAt+I8ehybl0AUIam1F/SQbMlVrwYx+c91raQi6vE/qNDywM+eZe9RwvBD511p2aP6xTmZkqVODZeA/G2V0ATGHiK/8HVR5Y80PPB85PfkDxL7Ov3REaeyhGSZh5sy9YHc6yPYutfgiBtZtJtzJ0C3vweevGdRMt6Cwf+TgY+h+Q8Syu/8BJve/AxFo5C/X/M49Q5ppf2bLpXkjGq4qAx2e4TazXznnqf1ZC1NVKHWvlpRfK9JkXOhteK+ysJksfHCjd2/n0l60vnTyyZeK91KAvwreqMznMI6Gh0f1/h74Fs9+3OjI4/katG5P3zlM6A5cQyHCsKcsfhw/t86/QS63NVXeBtd/X8u4PU/i3OlxbtkNXqxkD/sJ/qPyFN07vUenLf0dfND8g1rZlHoU94CzMw10awQ3radHbkdObT3zAqtMZrnN8Nqy3K0P4cymqH6lS6nmTXjVfoT/5wmtUfChBlT/9hFwBslUwt5n/8AlpEMKUtgQLnHdXPJ0DMCEWMwX6IjRUA8cfIfu/hFpwdr7Ygzhi7SfFaFqCAASdm7TzadhzHLVgIqtdxu4KfheJuz9KwQj2qS9gB/tiAE7qvlmSK+mskwW91QxdPYDjbRSO+N9HnVUICb/ArikZuY0ilyG8wem9/rpFTV7kLcEPLwhP8GoxPp/WPkKZaqcjBe+CX1bThvO/Z/LbhCatu471vjdevAABDenYcSr9Ke78+YlNua0KaQPY0Xtj2TId+Ywe/dy2UNaD8MBtH6f8XzwOXwyTMnD6/Q58374I/8Rw+WGK7zI2PZqXR9cdAuAzyp0Cm7BL8GnFJ4P7AI0Lf1aa844NoHq0l4nXiqqfDxtlcdjIc5bfhTY711nLhaTfQaOf+HTcSTM4SJXRRor1qbjQ0G4mdIxyMKGmbjpPdRbS1i+16tc1vBsb2fqPtWXDuPsT8GHbBl1baL+V9YoFn2PWmAUo8TgfDHFg5t2GkNai1zert83r3fNqZ3zwwxT5zMNkHj9OlW9/Dz5+Fr169gVK3vU41T+3mc37dkCSMp0IHD6ftIEQxTyfr3dxl1eNTX/sO6Y5J8xlwb4v9yBU5ep9DFWzxTpqVjmr3VzbSTRwt/ZLu3LhFSpiMQ7gBBGbCtWpH/Y/42slwCijyh+NEzNz/hcaLmWihFnCM0Oonzl9BGZUphkOpsUXQSgcReGyhdNL8D+DkGKefQWnETsFvw/DaVXX/q5nymiCVnXYAKZIZardIAUiuHDzQdRxuYFTVf/WMklsUGzzP3cJU063eXJTNcHc7GHXhEBJ8MWz1OlcNjugn5uqYxOZFs9R+H85RcGHK1Q79kmclrtPH2qA+dM3GW2iFslyqBDAy+2YZ0Bs4jDQz3gD9msy4f+lNifYaEWOXc2Tp7AQKLPYBzB76rVS+T//L+VnFphe3MAnbfMDx24Hq3jkGz+miPJlm6EMH8JprUt2CeF6/19KP6x93SL/T1s7t6pFXxPJGy/elMFXT/PdzdO1hncvvUYZONk/jFP1Bd7gboeuzTffztnSqnJftOVEucr0S/20hS16/ffFZQ9P7NWV68duJGj9/uwvKAjeF/n2B5SYSlL+fr3x0H7Vu9GG1LERAodPkwaTZPLReyj3NO5Ka5r0md+7QEdvuky/UO92nHqKswq8N7MI4tqF2JELuPuoQdkv/xNVoNJ/lYt1Ool6AtX3YYZk/6zR43wSBwnCW4Reox+xXxn7o0V8Bo3LCeGA/tcv4lTW2RoVodurrFF93wxfNyw8lH0Vvw0ch2aNOeHAnRSH8PYjtMWC33hL8COKPRimo1U48Js1Sv7lW+S8bDEVdPT+eyiqPm2UYOJ4DMfzsct76UXsoP7FwO5793wRjLtDEFohOOJOoPw5nF09+1aLIKUVg3lSazzh2Pztf6LSkU9TbA3Jt1ECJzG/CmfdV3EKN9U8QvUX9S41/iD75u1SwhH5EDSKP8W455/9ZwjEDS203nWMYpvRBOwSGVLN9YVA5KHjNIYT1T+CRv2rf/xfcXoOB308X9HB+49TRrk7XK2EQzLHg/R1XAf06rehCXkzSKa/Vh46tq4/2lYoDPK1Rh6PzDw9RxX4tmoeyScHUdMAriM5diO9hKs0Ss/OQpNua7cKHCRKHecrH3okXLfBV/q8xPculnAJcPMVMlsadG196FFq1aNevFtp4kJhfchgO3Rt1Giv3z8CnzXsHr//5mWqvbCAq0JwYMJTELSsHsz7UPYXOARQeH6BQo9F1tZ0Gw4m3P0DegmHyErPfpecu95p4Zi8n/m0svXsMOGqE9D67ssOTgnjInEoI+qeFpB91SRdHQQOnyYNuAYfSFD9K8dp4u4gBAW+kuMIDeFKjVOFFOX7nMhRw/Eb/55KX8Fx6Mit5CBCQbkZoFH4SMx+pfPE4Efh6O/vZG6lRMQTbH4DGjb/1CbMktHWy/3DUHt/lk4cQxmcgCq8zMxqrXjBfmm+Aix6vyf4UWdbnYIftGyjcao+GaFxMMY6hLUGBMPx+Gfx7GObF15uwxHxOLMLh/4R14SUf7V7kzKI+5sKOGp+9KZ3qAIzZxQCYUxV75tX4fMF+kdxhYcNrEsXejvVRh9L0ixeMqNHwOxxx49zl0GTjyepxHfH7VaCk3HxmSiwxDF3XCfAGtIwXrJzX9N34e1WM1LPAUPgtnupWkjQNBzFh47AuRzXvwzivsLxRx4B/7l39zS9m4LtQ1hjf6DXyi2XqfJDnH4GHzrxZAprZfc2X/Qb97Z4JMFSUcYVOyMwlZ3jtaIY2Ecp85VH6OTxO8nAwarSi5cpCN4787Vkf6GV199jYRoGhuYPIaDhxHqWBT62PvjXnGyEQTfvxmZ3mHn31+IU2y5dG7XZ8/fbKPskhHfwtSZ8vEqXcMjgIa1x5StZlGh1L5zzcc0Gz5mXfmgqK8nadDNl4BPLOIbgm1x6+TKFYLU5+ZW1p/TXlt3skw9RDPx1Jo6rpTBWRVwCb+Fk/Ykn/2B358xmyTmk+W5wkfay79PT0/Tcc8/Rz3/+c7r99tv3simpWxAQBAQBhcADDzxAZ8+epZtvFr8ZmRKCgCCwNwi8/fbbdMcddxDdfj+0wvHVjbz+XRxuPk87FbEOpSZtb4ZLahUEBAFBQBAQBAQBQWD3EBAhbfewlJoEAUFAEBAEBAFBQBDYNQRESNs1KKUiQUAQEAQEAUFAEBAEdg8BEdJ2D0upSRAQBAQBQUAQEAQEgV1DQIS0XYNSKhIEBAFBQBAQBAQBQWD3EBAhbfewlJoEAUFAEBAEBAFBQBDYNQQOvZDm2Ltx6d+ujYdUJAgIAgcQAce2d+V60QMIjXRptxFwZK7tNqTXsr7DLaSZWUrmG1cN/3oat0RXRChcBbiZoUii7IWlsameT1AsGqVUqUzpcJIErqs2PaWhvULAqVEmXd790Go7obdZRNSTPC653p+pWc1QHHwgmqnumnBbzyCOcvng899GIUnZ+v4cV6Fq6wgcYiENN9RDQIune4TcWA9Hs0ipwnZWgEV1M0jRyNpoAus1d9B/a9ZNCkQjOiyNmad0LUHlep1KsSaZwSgJXAd9Bhz8/lmlAjnJpA6wvk+665h1shDmztgn9Kwiw6lSNtOkdLVO9UJ88xFS1u2LRSb4byx6GPgvYko3D74wuu5wH6AfD6+QBmEL4hYlt8ilGtUSNYNbLMQTxkF4DztC0W0UPUDzrasrDpn1JkWiYfXcqtUoGI+pcDn8ErEj0f35Ejm4AyI9220EoEXLlw1KqfBq+ydZCLVmxCK7JADtcr8aVaqHoVHfTcjAf2vNCEWMXaZ1n1bnOCKk7dOh2TJZh1RIQ9DuQpnC6bjW4NTSFE3mKZ9KQsWOIOrZKpkVmEITiMkZjkJ1zBO+QcVklOLQvjWKyAc1vE01SoVTVPXXg1WgaKyozRowpUbieSpl4xRNFKmBQOdmOEZKHGmWKRlNUMnqHi8ILcU0JeKIJxcJg46aZwbE81Ka4rEYNHEQ9FJl8oua2SglsgVKJ9EOhJp0Bcwon0IdUQpH01RtdrfR43sTO1eUj8dQJoK4l6pyh8qJCKWKBUrheYY76VhUyYB5KjrQbtFUpgi7FCcj3dYu2pUkBQOJDlMlApPH4iqQsNMoUQoYR2HKiKdzVDYRxzRiUTkVo3jOJKucwvMyVWsmhWNaeNuNdslCHzGeivZojFJlD8HuceqAx66j7xiLOITISLxApjfO6jnqiaEfkVimjfGm5lEP/OXRgUXALmPeRFMUUwqcdfhFHWb/eJYKmSQleL7F0qSmKPOUTrMkzJSxQAgaZx8yh+qZKNwo7L7rU7ebpALzENBShn9cHeUjHOkbK7iex1yGfQw1rE49+QKyqOe8lqDpjsQpX/NKOmXQnqZCFvwHAiCvmZpZokwSPAP8LF4wu1tY890sJhS/bDYK4GFZqkHITcHtoU07ivRby0xaLU9JbPSYtmgsSUV/0Tbq1DDaMZatimdOBdYG6vd5cSMP/p8pUj6dVDRH09WWmbofP+jb5jXiLexnLTLamql1/T7g2J17mZ544gmODeoidudeNrO1upfn3MnhKXd+RRc7f2rEHRw95S7y9+VZd3xg0B2bWXTV1zPj7tDUvFf/vDs1POme88q550+5o+On3SXv15W5SXd4ekF9Wz4z5g4MTbpnFpfV96XTY+7oqfP4Yd6dHhtzT6nGVqflcyfcYdR3kR+vgMajaAsfl1HvyOSsfo7WTo8ddU/Mc6Zl98z4INqcxycUQfmjgyPuyQX1zT134qg7Mbe2HS7ZTufdmdERd2pe08lYjM5cxM8L7tTQoDt68px7UVWx4s5Pjbrjp88rXBiniaMT7iyKrcxNuEOaIPxw0T09OuDSwLj6TZWcn3JHgeHKCvo+OunOeoBdnBl1B4ZPukAFacGdHtb1AS30cdRluHan3UX35Ni4e/q8h8XiSXdk5JRqt3ucFCmcls644yOT7pyilekZdtXQXjztjo9i7nh9Wzw54g6fXFRFNj+PVHZJe4jA8ePH3UuXLu1hC5upmucN1qOe4DxB+vKLpdNYC+AXer4tu3OTR90RXgBLp92xkZOunmGuu3hyGPx00D1xzmuf1yHzrnXWJ7c7Ar4wNeutXVfzMWYNF89MuKMn5lo8rN2rfnxh0T01Ogoeo9cSr+2h0RnNmxam3aGj4+4ZxTAugq8MuEcnzujfLs64o+iHD0V/9JhvDXv8Dbm6aV/pv5bV2hzz+DiKMs/1eTd/Vnhy4vUPzDQ7YL425PE8j5+CV6nlrXgcxoS7sx4/6NPmteIt3J+N+b6GQv7eGQIs17B8g9idLv32M6v/3PoJ9dtO043Xr3i5fcqdepnqkRSV1O7WhsnNpkQeO1n+zluQUJJy8FVTX/EnaPCOE8kyyQzFKKfKoSRMck4kp8xznMxaA9of7eNm1ixo3SqUirDOHpqwWpPCKQfauBw5uQpl1zhb2VQt1iiaLWoTXwDaq5IBzZtFpYIFR3rfpyVETE5NaXWgNWuAnnJMaQSxf6JAPEe5qP5GTpBCQY9YTeLav7HTLQagcfNsC0aiQHns1MmqwCcsTcU8tGRcyq5QoR6lXDGsTSQw+RpBbpEoEPDa43zQMhUDMRpDs4pE7EPLoD9ZiJFTSVA9kaeCB5hhGBSCVstQ2YAtfNDSqio2TYRxcACfd6FdqkHbGc5Ck+lhETKg6QNWqH71ODEhOlll7ObjRUooWoPQ8hWIFXtmvkhwlmmZYowwMqjd+hbmUbsZ+XSQEbCrVGliffI85hnSl18wf7AoBr7gzzfDCGJdYYYGgxQET1JrCb5ahYpB46NNb21Bv1/EfMyUKYJ1kuizPrldO1mgQtJbu9AqmSGsNWh+kxVouLEuvSXZHo0+fMGp5qgUzkE7qNdSAGsppDgPs0eTgilomQ1vnYF7ZaDVNxTtyAGewct7/dSA9s2AFkvn6qbdqRX6r2VgEczUW36sQWCnMAR1yq3C920BL6jVgKsik/8KkKF4PHzgGrCcePwUhSkQDFEIWdbjB73bvFa8xSbLQn+YaEkHAoFDae40K3UKwxyopzGYQiMCNb03qdmhNgrTYeurBZW+oQbbgcnSgbnRZ2gNFso8fyoWRkwzAObCBRtY7DhJ1HKqYMbDlos05AUIRa3nHXPIqVPVhAmx5dgaomgcDvXNGlXtBMU1CShgUcMytG+FUuHDNOpxPovpgaq/3S+YTD1m19HSqo/sBxZAGb/6QBimAjBZ5RMWS5Bf3KlVqAETbdQvbYMOMOEwtw1uF1D6dZsqBRY0M3juqdzNApUg7KUMh2oVCyaIcKt9y4LgCpMh08vYNmFCVb82ai3TxM7bBbOs1ODvlqM4my7ZlAxBsREOYxy7x6nVOarBTsz46wSmB3NtONiAGTYIM0z7lWaZFhkwi/CYb3YetQCQDwcaAadWpQbMgf4sWo9f8NxJagkNmDh40Tqa77Cg4IlkTZhOG4kcpQ2WebDeYAosVCMw40NMWmd9crvRFr/TwmITAl8mH6RCOa3XXNdI9OMLzDsjcBvw2CPYUYOaBkyGLAhhI9paM7zRssFLvOXO/MTB+m6vnD5Db2OzBsTU3hapm/b11nK1isNZ2FS21nmuTobiz5Yn+HlU4x+rUqBsGj7JiThlKt6BLt6EG+C1XtvE/BXuLhF2bOnHD/q2ea14C7eL94OHex+U5fF1hMAhFNIgTDUCFFbSBa9fLMQgfBW89cu7wRD8GfRXTHjlM6W/dWrKeOHXzVD7tFCnYyoYTd0Bo/E5EmuJAmEIXDaF1I6tR4JgYrLPhEeHn4OFFz6F1VpzXjvMe9Qus8X44Gdisp+J1y9mdtDG+V97tIhH8EcBY/Ud9zvz6BeK/3phZsk+Yu3v4MhkKQaGBE0aC2mOVaZiM0VZSJTYxOJFAu1goQ7/vQR20Mw8IHi2O6I0mJGoxoPbMyCwcbIxBuiI2nXvvF2ME5hWrgpht1pVf2p1i5rwowt0j1MLAIwfXpqYBquTl98jGb81gXlAY76FedRVq3w9oAhYjQY0xZ72aj1+wfyhQzBRGnIT/lBqOWBt4W8H675QDFAGPqisIWK5jYW2ZjKjNpT910kXn0Jdaq0Z4GDgSdiP9Uj9+MLaE+rcrqH4pRYOWiwC/EELOLp6s85Whtbi79Gm90jxFc93txsz/r7eWqY0VeBs56/zeqNJ1SSwsqE5ZHyZ1WAznIc/aQ6a+mQ2T8UiNH1B+MgCDwdl7Y5NuH/yPIDSfflB3za73gGtHq9TF94ZO+YteI/UQ+2Ne3+g5ZfrBYFDKaQ1mqHWKR/e4TWxMLFGkZg5QXDwpatVL2XWlEHI8KUo77RmSxDDiSQ+GKD4qmJQPqNBrUrQgukwG6ZKrvd9SQ4YehOc06ejUYW50QZ/sSwKhPVTRWHHjpmZbaTF+MDYLWjOfD7oCVHrs8UGmVYbC2rWqQyziy98tK8L0Tv7kMFik068ozb8HTWbFRi7YplCcBo28E2ZQlloc6BFY1AcHLiw2UTqVQAtQLmOF5EikLGFXKbMtNpMoPu1C+3i5WFCOGxDCMZdgEMzsO0eJ48ykANaA6DVe4HZZoWqDX4r4mBDiDVwXrJrypyF8xVK87i5edRqRT4ccAQsCAnajIaOrsMvmD+YmFeGjwdetLWQr9EJKNN8s1qgejRLCSwRFtIcCBulMg4QpLjUOuuku10l6EAbnIODvYXDUupQVHfqxxcsajSxLloLwIJmOYADR3gA3lEn8D9vfTexoIOtza4WFDdz/VADwlwIkp5aemtoX2cts9arE0PgU4S7SZPrwdr0BT+7gsNc4QKVM9BwoiMBq82rtSDJzEo1jmKW3sD24wfrtrn6HeBVuue8xcJ7A6r+9lxqNSwfrlcEDqGQxrcxh5Smh9PqHSjvcjpMhGx2g7ZIyRGswbLgI2Z4Q203qekfoYEAUsyWW/d9aW2cx2iQXR13x7Y4lACT7WSM0DQ1cRKnlXzfEwg3GZwwtX0hAcxe5eLdNE6XJrJsbuhifIphtDVxzOz4iP2GidvUlcNUkKYStIyKOeK4elvrpRqnpsWSDZJVomwNPnO+eYY1aTixmq/HoUVjYOHvASGtBiZpQIvW4unor+4utH65LFXgF6NkXoUtNFcGV+5p3FoqxZ236+BUasPiuiH04vRXrsJCpB6X1guB+wdNp4aCMfHGhcc2ncde2hsMCM2K8bMJBH5qTjqtTMCbnkeaDPn7wCPAcwjrwPcN2oBf2DC1WWrK4aQzeEkYpzz1uuENEPxSizaloEXjWch12uU81eI4oehNy77rs0uj1V7bEfAYg8qtTWPH/GcyevEF9dwieCmo1MQJ+FIog9OXvIS7NynQ0HtuIq02DV2u/9/ar7O1Se6mnZvvs5a5TgcY8l6KsTALoA2bQsawARODv86bWL/M/BVs0LDl4M+n72nUWkLtrsI/ahOp0mYqLHrzg35trnoHdPL5deqCjXuHvAWn78tEqSQGRNKBQeAQCmlYnuznoVYpm6vYH8GT2GCyMjtNn2pxe6ZPONMbwSoYUoKKDRSFSjkVxfFw+K8l0yWy4RQbUY4A7JvB/iSaxfLLHPxLm/VgfsxkQmCMeofH12dEs6aaTAEw3Iyd11dEpGqUxAWYMdAYwtUg6SZMGnCwj+NQgA0GnmcBRu0yoYkyVHHP9On7y3GbTAP3C8ISPPCjcN5fm3DFRjZC1VQU7eIofQC7zDToVqZXz5SpCmG3nMsRwfk/ClNBLGPiCotC23eDhTR0MgCBRbMHdsqFj5aToIzvYxeA70fKphzaSSSzVA9CcwC/MIOrB3NlR2Yll63SXu5Cu9jdZ/MhKsG/jGlPFAI4aJFGu6vHyammcWVAEawZycCN3QkLtKOvGBcLjBxKUMKbk/LxOjSDfMQfx/tDeezI+YfNzyO7nIAGsrc2lZuWdFAQ4MnMWjCvP335BdYq+EUy4VCG1xau3qjG4bfKKjOVAnBcB18yMpT0WApr0sx6AFo0vdrWW5+s0Qq0NFqojte2p+UPYR0mGqxNA5fonP+wB/TkC7yWCmFczcP8AtfYVGJUwoEmJstSZk9/Y6pNn573gmqzZfq0cIWI0XFtkddL/Y/26/T3lmto77uWUTqKtRnltal5WRY8s4zrNDT/bVtHwgnwKFwRFI0nYe6sK/eTMO+8vY2pL1f6JlKl3O/HD/q2uZq3dPL5vnXtAm/B/U1UNrLw/10Fqny5zhG4gY+H7mUfpqen6bnnniMcVaXbb799L5vaZN0mZaMIiVIHs9hkies6m427i2J13PVWaB2GuK77c10Tb0HOTZJTxHj479fruj/7l/gHHniAzp49SzfffPM1I5I3P7UMfKBYVuibajjFDH5kYp36Al3fvNf/DyywZMNwH0j5m9jrv0/7owe4+zMVx3zD5cnrzrf9Qe1BoeLtt9+mO+64g3AFB9Gx+Opuvf5donfO8x0cO+ruv9tR6euycBDXUkB9fV3SvnWibWgDw/mcCGhbh273S8Bvx8IFx0r5JunAIxAKse/YBt1UWuS2m8IGua/zn+FT5qRwabgIaLs+kHA3KcH/NyMC2q5De60rPIT3pBm4Pb90rXG/au0HIRTkr1pr0tC6COD+PbjKSDokCETyuNdsg77qE4XZQ+LoHUbcY9mhbDAltvdzCJEkytsrKqX2NwKHUEjb3wMi1AkCgsDhQSCQxP2DycPTX+mpICAIbA2BQ2ju3BpAklsQEAQEAUFAEBAEBIFrgYAIadcCdWlTEBAEBAFBQBAQBASBDRAQIW0DgORnQWD/I+CQ3Xnf3v4nWCgUBAQBQUAQ2AQCIqRtAqRdz2JmEP+ujBvUJG0egQbuQYoh7JQu0SzFKBhfH8M67odLVjY6Xrd5CvrnBG24O82nrX++vfqlSunk+ljsVctS7y4iYJdwOW0Q9wcihBHfIWiEKY5LzLY6gx0ETA/gmiFvqewigV5V1SQZ6dru17tujQ7uZ0PAdnPdTPKjIHDgEBAh7RoMqR8Tju9JlLRJBPiSW9uLv4ciIQi5DVwM2x/DtXEGN9nS1rMxbbgAtB13b+tV7LgE32i/40qkgmuOAE4Alxu4lgNh4iyzQKFimgqNrVEVwGW4jUqqHelja8Vx+WyRUnzDrSRBQBC45giIkHbVh4BjwvUOan7VSbmeGlS3lnuxUZnuYAj33QX690BFZNCBk/tn2qVfVPiadqzWXap1a9U4TdHMbg2x/Z87GKV4uD2Y0I4AACAASURBVNkKabZpglkb54ej2nShjoyIvFK3tqq/205DUkYQEAQ2QuDwCmnNKmWTcQQHRmzOSJJKFkPFKnWERCkWKIXnmSoYFWLFVRB/ksOgRBFyJFE0lfnBLiGIbbq927QRaiQYSFDbugYTGEIglaDecBolhI+CwBAF003nqGwaqIuba1A5jXrxPIIYoemqbwCFIFdKgzZuE+VSZR2uyKOvrfI3cXlhgsqKnyJEVRj3cHG5KO7MQVU2ogyk4ugjgnBG4gVcJKmnQxPxK5NxDkGDsFQc3sj/Qf/s/Q0asItPoHwM4a5i2ZonBPSjDXGMcZt4IluA6S2usEpXEJMzn0IdwDiKcDebUfX0HBeEnUEs0mAr8DKH5/JDy/Shk4NW+4ITLnpMRhMY4zoukkUomxYdbKYMUNgLzcUd53GM4pZ4vu54vTHonCMqTh8GlEVGG+F1oh1YtyDlscY8igN3NdYVj4g+84tDTRUx/pkCwgEhkH2E52gdQaPTqIPHE+ZNy6+cp6m8U1fN3n3zBeNeQuQBAxuKYBixbautUdsUiY6Kg4us1RSFk0WEbcNcAD9AFFyqFzBXwTsUX8lU2vOhjrBniYpnJkW+fJJiHOIM8ymOOlrL3apQxuMDEYSj4ilplVJY61WyKhmERNPPehKKuJd5rPMY4iiFER7JX9sOQhMleX5iPYQj6G/N52lMB4Kagy908lEO29SXPvCCTIzrYroL1KqqJ0HyUBA4oAhwWKi9TE888QTHRHARFmovm9li3efdmdERd2p+WZU7f2rEHZ25iE8L7tTQoDt68px7cYV/WXHnp0bd8dPn8QlpedadODrhzqLYytyEO3Rinp8iXXRPjw64NDCufuO0Mj/ljk7Nuysr8+706KQ7u+TlnBl1B4ZPuue51MyYO3qKPyH/uUl3aPKc+rw8N+mOTM6iVk5L7umxo65uasGdHp505xQx/NNpd2x0Ruc7f8odGUSfZj1al8644yPIq9rlOobd6QV8vHjaHR875S56dSydHnOHQGd3Wj53wh0eP63rXplzJ49Oukxdf9qW3TPjg+7w9LzLEKyg/FHQc3JBfXPPnTjqTrQI727N/95vXFbcuclhd4rp58R9BW3ctX50cr8UtsvAf2zMPaU6fN49NTLmnvbGYmUeNGJuDk0vehVjHMfH3dMX1+vn2jkyO6GxXVmcAbbTrjetvDr5nxV3YRpz7NSinkeYG0PjZ4BT//kFAN3JwSF3Ug+gu3hyGPPmhHtO0c7zoN0PNQ9GTqk5JUkjcPz4cffSpUvXGA4e9yHF/1p/BkbcmX4DtYw1OzSF0dVpeeGkOzqEdcdLSK0nzGO1npAWpt0RrH1d1bJaHxM+81nA/JqYU3NthfNNzKq1oufhsMfrUGZiCGX0L0tnxt1hnw/w2u/BE3TD+Bu8amBw3D1zUde5CP45pBbneffkCNa8x1y47WF/Xqo5etJVKw088YTHx/rTx10cccfPeIv1/Iw7ivXQWqotYuSDIHDtEGC5Rq3t2+936befWf3n1k+o33aaDudltiYCGAeg2YlpjyYjUUBw8RC2kRUEWE9TETs+g4Vyu0KFepRyxbDSkhCCqBteSKkATAqseVMJGqtiIEZjyKSfNKmMgOZJBCRnJ956Ik8FVM/JMAxoXWKq/prZgIOvpiEQzVM1zJksKqFsCoGLOQ+8r2C6QF6umAPAIxB5RhGDthDUvBlJqXw2PtvYbRaSmlarDMdhBAxPqHaDCJCOgO1haLvyRQpCU6SCmfMv2KYHdLR5/UD9bVO1WEPw96KmgQPDl7BjXo82gtasEUPwcjj0owbWRAUQED6n+gfiHQ7H5TXa0dKqj/3Ghes2oc0D/Yq6eg0R6/NAxqZyTzodqteaFE45VEzmyEFA+6zqcAB9BWVqkDBG8LuJjA9Twxs1B+NYCmWoZlhUTK0zBp1zBLTVGwhgHyhTCprVdKVM3rTSxPLfTo2KFUS6ML0A1BHMsVyIguvMLxWUOpKlih5ARXM0k6e4N494XiDqkE4W/JdCke37ILUplU+7igC0p5U3Vtd45VUqQ5uW4aDevRJrfcN1FZg9iLWewjqOe+MciGUo4/ELC1rVQDKDNckJ6zsZoQw015SMrarVqtUoFM96cyNAUQRHbxYbyBPCvDWgrdcTKsQ8SvGfzaUAtPUJg/MGoKWPg6eYWFEpaAprFPBMrQEcpgnbVa2BB78M2BZZNpYuzLj5SghrkQOz96MPbNUMQjPv0QQrQToCLeDmyJNcgsCBQeBQCmnMGNi5VvEYpABMYDH861Qg6IARRrznTg23gcNMF/W+E5hMA2wxyUwTXDSg3vY2VQos0OSJ0nktAMDht4QXecVwqJa18HL1pAvktiwID55pLJbOUB5MM45TXaUsGBrX26xR1U5QwfAbRZuWQbAqeEJZpkV3o2aSoeym8PWtNWBmi2phEjTVYH+IZv2eBMiA6RW5KI8XRL0GE2iRhRZQj9iBRq5Nn3ro1KlqwnQZ1XmYoUdZOmiW+tJGjTo1jDjluQ9IFugJw5Sia2hAgIOQ5ZOjs6z5u9+4UBOO1GDsaa9us26hbtAM4acnnUqoAw2FNDkwDZVbUpM2HTV52GCCLjoZCLV4KSr7D4+jSQkIs4F1+smCceccUbQ5NjXSJUqUKp5Q3NU1s0r1DmwoGIHJkudb//nVhAkVdm5vrJtk4oUVT3svLA8PX1jnvAG8xDx41uAqD64VAiEKGwNEb1zpIGAQGzVvHHuRpQ4OFNs8pyNPgDdU3ne7acN82h7xAEvsODzCM9nPw1kdXM1iYsMXr+inTrNBtpFAPhhMHZhg/czenOxFUq9nIfSh1Tr8Q4PghdiWYePQpFqxSBXEDLbQVt3vSRA8DZulBMyd1XypvZlch76mHYRA57eONnbiZ9erE/JMELgOEDiEPmlayxKB30R3YkGHBSg/sRAU7pQs4OdkwZ9I5eCdITMmq4yrF7CDjBvKd8TBC7taqMNHhE8egkk1IOC0mgLDrNvwsdBMOhDNUq1epAiOtEdTFbXjdLgN+KG1i0DgaMIvDQ9W02fjxU3wNWEuxicZQ/jsc1zW7kTwvauHNp4ThEdooqrVqvpTbzSpqqTOjgQtjmmsDfq8Hm1KkwcGrHsGPxPQFvE5LNo1oY1rM9wuutTX/uPC7TaBu8bEQr+9vvahUwlOgTAF8SLjl0k7aU2ajXGrFSrwRUtRBC83lrXZb7BEGcKtHVsYA+QFlk1oIJwmNIV9XiJKiILfUPeruf/8wjypWRSJGZp0HILw54BCitsE1vpXxg154+1529Fh+XhNEQghVmWexgbbRAxNFLEx89fp1SEumtNrndd7zbRgIEh4ghzo2JCUOmWjsCAY+IONQLnp09xVUH01lR9u3oFfLzY7pXKOYq1s8P3EhsmsZSkIf94I+J1fVX/6fK03183ipyRB4PAhcAiFtAaZVkhpplRq1qmMlxybv+omzAHKLMbJgdbLwUu+LcCY0LQZcOJWOdSuFi/IYplCeNkb+KZMoSy0OWlK8TsTTsMNG8zNrwKan3IdprEwK8zKpJplDQuYmYEXbwNfbcuCVsTgFlTS2rwE9qPQpjQ66IO2q2bhSgrO6p1kjPhSAHawjQDa9bpimxWqNsDkWNsVwv1LXt2sMSvm28yy9RjH/5vMlNUDhxpVmIHt9WjTAmTEt0eyJstqmyfBmZVwi26vk/qNi67b8IVl7rd3FYfTh04tTMKkmA1TJVduvQz4jcSaAwcmpQICJmbjENp4HHEyslaoUiSrr/RYbwxWzxGGFIJ9qkC5GBypSzyCaxPnCbUmnEU1zCPWefSfX1oQbc1FNn0a3uYA1as2fawxpyrQeiZbAvra9uXJtUMgEIH53DpP83OzNLdwEVdj+G4MO6MpiM2FbWNResnBhoRPPLdkIu95ABO+2fTFITxslClbwg4KOQMBaNP8KsAHK4ohdSeYJqG5tsCXLJRNejzGRp0tsQkaPJu1aY0qVWDyLGQghLGmreVagDot5n1NCkCLl4dLQByuBXxQoT992PR00octbFOpwCUJAocLgUMopGGAWQOm1juErHyaShB+lKDjaazaU8ChJjtRcLJKOKkE3yzPR0hp0vCyz9dxginOUhib0qChgdBjQIvmy0uwN8CswBVAu5SDjxH8TCJgXvViBu1qpqO0UFEwNq9hG9otTZ5JhXwDfhksGHYyKWhaCjnU5Wm71BUQ7Ze47p/H0CAoFmGGhS5H1e7AvKmb5TqyVIIQqWgFnU1NqKaCMeJPEDozOO1l+wJfT9q0Jq8tVLAw2NbEsVDRErJ07b3/7jUu6DdrDFtaOWC1pq9ddPJpSzYDhxJZSlh5jJGHBfJh2NDvEgUyOC3H5VhIg3mz2EzipGybrJ5jsGaOsGAPqySk41g2QxgsmIO5DhiT8BJTH73EZidOdiVLGagjvNHoPb+geYQBvXXvmq0AiHrmJX8zgY4gORD2TZhF1RSUtD8RCGLDkkjCfYC3cruTDGhmbZjLLVUdtPdlE9pUNaNXJV4HzTLuFFRPme/ksclhKnizaFG11lS/NKs5zEtLfebE87W9alqPWx/sWkkJWZwa7IeGTVSQBTVsEDXLxOauVFS+tKqeZoVy0OipInBGY59e3jD1p8/AuoILAtwzVAIfKvM6Xo8onVP+FgQOFAKH0CcNqnj4aqVT8I0AswjCYb/Evj58ZUOHtoJ3mvFcTh15jxbxIofPRrpcaL8MWUiDwBAolT0NETMd+F7Z0OD4Wo1AnDIpCHcwARShmeKrLAy+SZx3sSnsKCF8xQvQx8Ffo1iMKwYeSubhfA4/OJgxQ2gjkoUgqLR7cPJNhXAdQ5RqMIfGsFNl0ywoh3UPIhh86TiXSga0RIkUnIwh3KGP4Qy0e2H+AT5w0SS0fDC9wTTHu/xyQUsm6voMp0RWEX5tcTgol7I4sl/gyigNZhvjyvvRpjR5uA3cUK3rQwxwemfa+AVigrlGcixFQIhNRygbhpnVz6yL4O8+4wJxhTWGSUU/eDWkoqDf15504jBBHbXl0DoOg2QyMDfzwQGYlJmeIHbndSuOwxyeVIM8gXqFQt7v3EbfMeieI9xvps1AoQBM3hjfXMnCONVBL64igAYMQwafxByFcHVGtMwk4cqTotbY9Z1f8Jm0cF2D12VtzlSNIKk2w7iChL/gJQYn8HhOzx2dQf4+FAhEwZsS0NhjjUMVjLWcxyEZveI6+x/AGi7hmp1kpKI1Wwb4U17PrCQ2FbU0rucosodalIo49KMSrquJ8SYGV2jkq6WW9syvlwU4A3zITIOvsSoOc7rADCZoUD4DXRqu+ijAfGDEcV1MoK7lKrh25MGXEogSEkCZYLpIzBLWoy8Kk2kkhStBKuCtcL9IJALeBrOzh/JZEDjYCNzAx0P3sovT09P03HPPEY6q0u23376XTUnd+x0Bu0wJmAWzMHWIdW4XBsvCpgGn38rlpKdl24U6D0gVDzzwAJ09e5ZuvvnmA9KjLXSD70nDCfVGy/dsC2UlqyAgCGwagbfffpvuuOMOwhUcRMfiq8u9/l2id87zHRybrq9XxkOoSesFgzy7GgjY0IKF8zkR0HYFbJiulB9dVQS0XcHz4FTiQFPlBNgWIEkQEASudwRESLveR/A6oj8Yh3/YdUTv/iYVh0ggpEX3N5FC3VVGwMLJyWi2ietgYle5ZWlOEBAE9gIBEdL2AlWpUxAQBASBa4CAkYJzfuoaNCxNCgKCwJ4gcDhPd+4JlFKpICAICAKCgCAgCAgCu4eACGm7h6XUJAgIAoKAICAICAKCwK4hIELarkEpFQkCgoAgIAgIAoKAILB7CIiQthkszSyFb7iBQqlaz7sU7XKCAjcEKFFxVG02nHc5NihuENofCbFB47i3zFxDTR13myHwd33ND/JAEBAErjoCHPs1RdWr3u5uN4j7AcN72499x2N3G0KpTxDwEBAhbbNTYXiEQvUyYl92J1zQWkHA7xHvclT8HEyWcNv2Lt5dhVBI+VTRuzW8u/2dfI9Qrl6nXGQndUhZQUAQOPAIIOh7Kr+W+121fne1v+s89qp1RBoSBLaGgAhpm8QrEIrjZnmOP9dVgGMnWrHWjfjqV9xib/QJtr3J5lZnQ2gnDoys9XTbqqFPIURcCHGMvT4/y2NBQBAQBBgBq45IF7vPgTYNbnf7u81jN02IZBQEri4Ch1dIsyqUQZDqIEIvGbE0lb04mv3hR1gmhOapdUlpHDuxgVBLKnKTn6opMmAaVQkCVikVR4glhJdCmJVsTRtBTQT/jhcQ7DgeoUTRQkaEKyokEQ4KMRvxJ47wKvwUgSYRDiVPdQQ3TiG8VJaD4dk1yqNcBOGdImGUL5gtAa5ZzSD8VAR14E+q7MX2U5TohLBCWfyWVfEsLSogrEzB4p/wPBKjbDFLKQSRj+Bz2g/Oh9pZMxhFW7E4Qk4VELZqP5lzW52TD4LAPkNArf8oGdgJBRHqK+vHotwSmX14A+qopQ1K5EuUSSKUE9Z9PFtruVnYiOyRQKiyaBwbzHwBJsgkAqBzQn15hIPC2me+FE8WyfTlL8TITIE/RMCDIgi/Vm1y2GKEZsKlyVYFvCWebsXs7OyCXc/DpQJtob5ktkKtWOjghWGfF6KAVYgiDBUqRVrLA5tUyTDvAQ/kuhBSj9lyz/Y7eWw/3ok21sOnk375LAjsWwQ4LNRepieeeIJjIrgIC7WXzWyx7vPuqZEBRVfrz/C0u7DSp5rFk+7I2Gn34tIZd5zztbKtuPNTo+40Cs6fOOpOzHkVnDvhDp2YV7mWTo+5wyfOueqXizPu6Mgp9zw+Lp4cdgdHpt1zS15lC9PuyOiM+s11l925yWF3YnZZ/7iMdkdOuote1vmpIXfs9EXvt3Pu5NCYe0bVs+ieHBlzZ9RPy+7sxLA7Nc8f/fIraHfUHT25qOlxL7ozo35+lB0edEdnNAXu+VPu6LDXJtM9dMI9p8hZds+dGHIHxmfxSZIgsD8ROH78uHvp0qVrTNyKuzA9tJrPDIy4/hJbS9ycO3EU66z7h3V4w/zUUXdwYtZVy39lwZ0aHvXWP6/nYcWbOC3NTrhHB3XdK1yfXwacYGF6GOv+osrHvMXnO8y7Rk55/AA8bVgxk14J/HQU7XpZua1Bry0wC/C/drmLM6PgXZrpdfPA5bkJdwh0ab6y5J4ZP+pOdvDUVe138Fh0oC/v7I9Pr37IM0FgawiwXKNkiNvvd+m3n1n959ZPqN92mg6nJs2qUuXVK6sF559WoE1bX5ZumTxbHvh1KpvQZK0TiNKsNygcR6BzrhoBhguIqed7r0XSWYqHdJsWfMMC2A2H1dcgdrcRMqtrXf3512jepAp20DorB1u3qakUdByUvUkNZZYIQkOHwOId/mZOPUfpGoK5wwmtp4UzEEYAd00Borcj5LlFvOd1EEi+iSDNcUU4aINm0CNb0yB/CwKCQA8EGlSuvLH6+ZVXqbxFbdr6vAGRJxLeegxAkxa2yeJFa5vQjcehFdMrPZRItcKxWbUahcCT9Brm8lFqgk9xCgYC1EQFzEFC8K0tw3qwYUJbdQcaOY91cFuaV2xYkjp5YBDaebMVhzYEbVqIbA7gvkHaFj4b1Ck/CwL7BYHDGXEgaFBoEEPwbscwDBgU3lDyCEF4ClGhDOEJpgWqV8iMQPBCNfWeI+qQAx4TbDl9gfF4Uhnz0WBQST0q2ZCyguH290AIn23NLLsFqgDqs6pFylfqYMhNMsFfU6oWg1IwRSZTYNYwkRZzCTK4MPM5HD7IpR1KwFy7yjSryvkpCJr8z7pVZtbkOMp3rZ0tQODlkgQBQWBdBEIUNgaI3ujcEA6SYWzIaFbVuh5v4IxtPoJFif/VmuX4nSGjHdcVC9ZnQ47tkAn3hXjFW+PNBtlGQpWLZIsUS8YpUs9QsQjzprFuB/WP4FN2MNTe+MFfrIO1rVtBJw8kuJ5ASgN/rZEJQdGCHy5l1y2uftwWPhtXKzkEgX2BwOHUpAUTVChM0NHWEAzSGPtsbIJ3huCTFaqW1XUWdQg8YexCN5RXNsywlbmA06SZKGGTC7+1HBVKRcp0SF3BKPxIzAql7Dx8TjLkucChARasHAh1ioXvLO1CFTsjQEoLAtcDAiFsmvI0xhtCLw1NYHOV2FWGsOlI6p3LNpqrUrWq//ChJLuCa4SYxlCM8jWTSnGTMvBLK7Sc1TbAexVPwJdt8Ai7Ar/XNLR8iQzlCiUqpMIb89YNyFI/7zLcm2lS8ggCu4XA4RTSgJ6RqlDj4jzNzc7R/HkLgk8fE2A30jjlmQhWYeasU6UO82BsPQ6gNU42dq46oR0Ids3uOvE9CM1Zp2rfYftl5+60VQamzmqIMsU0xcLQCGLz6UDTpRKclKsVk2w2WxZrVAxXqMAHDTjBFJIrpshMp6nKmrUtJNbc8W7VTw7vnLfBhLfQpGQVBA4EAoEINkrWeZqfm6W5hYvUgCBibLFnm+cNHRVDlRXgdeo/6lizvJ6b0MC3Eg4lZUu87bTJrFThrB+EsFSiai5IhWJ9Y2qZT9naNUJlhvmgxS4Ue2wzC59V9aq0Dg1aJFugZDSM0/FgbJswdXI928KnFwHyTBDYhwgcWiFNLW4jRgk+FdVhZtx4jEIoE6RqNkc1mAjWldFQWRgnSM1KTbMpnCjN5ittxtnRmIFdq13xTnQiRxUm1Yjv5AFTRQDcTbM6+J0FLZg49TenXqQSTmpqocmiUhYnQX2eqEwcHUIkXhiltEWZdG8a+vYdp1LDaEcf9rSoAgFwi3Je36rlB0HgwCMQDFMskYTvqrEtpc66vKEfeEGcogxUwRt4pfLp7FKLLxg4vdks41S6KutQvZiHxl3zCRMmziLLa0gBxT8gLHmJzaQ992ZB+MUGcBWRp3VrVkpU86WxELRhpqnbwiaygk1kvxTEtUWWcqhDsqtULFvsadFK/drfFj79iJDngsA+Q+Bw+qTtcBDY5BnIZClUrmzIdI1UAddd4IBALI8NJqIS4NqNMNpfw6qiOSol0nD0jYE7OlB8wacM/m8qBaKUCOdwFxv+LdQol09QIg0mzD4n4QS0eaSPvAfilMuWKI0j9wWIUY6BOuMo3yFRhTNFSuMqj1Q5StjUby7hFvRizqQ0TLv5AI78w1EZV8ZJEgQEgd1GwK5QGv6uvmhkpMs4JLQOb+jbfhja9gyls3GK2kEKw88sFtRiWSBWoBIOAiUjFQhhELyMNBXzzJUIV3Xgio00DgEEeeOH9V6O6Bai2JBmcJ0GNmz5aol81qR/DFMa1welMnGUwyYSPIzvlFQbR0QeyKu2qhQycO0QfuvYNq6iPgotWgSb5mgFFgIId4lklEylTQMf7G6/s+R6vHNVC/JFELj+ELiBj4fuJdnT09P03HPPEY6q0u23376XTUnde4YAzBfNABinx17hVBzOx6Ah9PxY9qxdqVgQ2B4CDzzwAJ09e5Zuvvnm7VVw3ZeCkGXZFMAhBb1qq9jkQQDEzf2x675v0gFBYH8g8Pbbb9Mdd9xBuIKD6BhrRDrS698leuc838GxI2IPtblzR8gdpsLNMrR1mZa5xKqbFIA/XL8d8WGCRvoqCOxPBCwqJnBASivPcI1OjRqhyJb94fZn34QqQeDwICDmzsMz1tvvKcydhTTMGWyGYYfeIMypJc8Msv1apaQgIAjsGQJsgoTZMRmmijqAFML3nAhpe4a3VCwI7A0CIqTtDa4HrNYARRCmyswcsG5JdwSBA4xAEH6wVTN/gHsoXRMEDj4CYu48+GMsPRQEBAFBQBAQBASB6xABEdKuw0ETkgUBQUAQEAQEAUHg4CMgQtrBH2PpoSAgCAgCgoAgIAhchwiIkHYdDpqQLAgIAoKAICAICAIHHwER0jY5xs1qFjeGRxBXPYILHWOUKjW827dNyuKCx1LHhbGbrPIaZ6tRChdNVveQChs36QbiJYlOsIcYS9UHDAGEaErHPT4TjiA+rx+FZJf72chTJJjceP3bJYpHst7l23XKGAZl6rtAi5mlSKK8ed6wio7O9lfTVM/gEtyyF6agmkL4v9ouECtVCALXDgER0jaDfbOE27QtSiGkickhTqoZcnK4hd/aTOHN5kEoqEyGys3N5u+Rz6lTPlX0wr30+H2vH3W1H0QU+EYp2bo9fa+bl/oFgesbgQbWb54CuZrmM40KJRuIGlDaCVPog0gYMXzNrV5si/i/9Trl9tXtO/uRpj6Yy2NBYBsIiJC2GdAauAjSQFgVL0oThRLY7SIYsRc/czNVbJzHpgYuid2RQg6x8Wom4t1t3Nje5OhuPxBCoGS58nZvwJZaDxwCWD91i0Mn4S5ClRCCLRWlJtb07ieEbzIQhH1LFSOWZyiEUFJbKrTHmfcjTXvcZan+UCFweIU0BDvPIPg5BxA2YmkqrydwIY5cEIKailXspVixgbiYPrdCAONyllKJGMyhMUrrSOQqZ7PGZtIoRflPggMZ6wrMLJgx4nhmYdpI/FWZ8ogHWmiYlEcdiYKpy0JjF0e5WCxGsTjT6DVulRHjEyYR/BZBJIAqb7TNAmJqIrg6zCUpxObM1taKanY9D9NFmKKoL5mt6HifXKVTpkTLpIHvHPYpUVHC3io6leqwSZUM9xP94bpS0JapjD3a7zI39MOilsbLKF+iDAe7hzk5npUA7jwskg4AAhC8ShC0DEg2wTDWZdXq3yne1ARMbLTaazeYLGNpRVUZZcrLI/YuYufGsP7YFOrr2BzwhHTc4xWxBGV8HqTaR36s+SjcMrI+A1plPuQA7Cn8HgGfgjtHFPE7eyrvLCog9maBu6DaA19i3hTBZbmeW0NfOjjIezFJEZhwY0wfB0/vg0QTfI/5ANMSTZVpDWIOXEzwW7bONXTQ1Kc+eSwIXNcIcOzOvUxPPPEEB65yEbtzL5vZYt3n3VMjA4qu1p/haXdhpV81K+7CqTF3aHjSPb2wXnYZ6QAAIABJREFU3JVp0T05POiOzpzXz8+fckeHT7qL/G3lnHtieNw9s6R/unhm3B0+cc7lZhZPDruDI9PuOe83173ozoyOuaf978tz7uTICfec19zy3KQ7NDHr8tf5qSF3Ylb/sHR6zB055bW9fMYdH/HaXtMV9Hl01PXJXJqdcAcHUb+ic9ad6Cy3MO0OT8z1pHN5bqJFh+suuWfGj7qTcx5w3e2fO+EOnZjXlKyDxfzUUXcQfVNdX1lwp4ZB58U1HZAHgsCmETh+/Lh76dKlTeffm4zgG9NDq/nMwEhrDfZqc3l+2h0ZGnGnziyqtd6ZuK4BXqdquTG/8Ncer8NRd9pnYEvgA0NY28jH/MHnOe7FGXd05JSruEXnWl067Y4Nn3DnvWV8fmbUHZpa0E2vWtOaR61em9zHEXdcMbn+dLhME/idJnHZPXcCfRnX/GxVJ8E5T46Meet/2Z2dGHan5jvpXQHvHHVHTy4q/uTzTZ+mhelh8EavI538Z3Uj8k0Q2BUEWK5RMsTt97v028+s/nPrJ9RvO02HU5NmVany6pXVwvVPK21N1RqxO0BRaHfqxQhVEGYlih3sKsVbIEzJRFiXCkO7hd2d2og2qlQ3kpQI6Z8MaMsMjqHn1R9JZynu/bamSZQ1w9AseZaPYDxBkUZdlQ0GAtS0mmonGoLfVzlprCm+5oFtIvZmjOIemaFEiuK+VWVN5tUPOukMYsdsln0/sxB23yGy7U0YadfFAvhCO6CgCECTFrYJ3ZMkCFznCDSoXHljdR+uvErldbRpwViB6rUcBUpY7/Gs1pJ31BBNZyiiFPjQPmPdN+rgCPAFrVrQyEfVD2AK0MYbDWpgDZn4PQyNly4CbX0+vtZHFGHfKnX4p3nFw9Bikb3JBYi4vvlalLJJrN516CDwPSuaIE1ikOJpb72vRgffYIYNNKlhMXcLUqJYwcGsdiannqN0LUFFOMZ55K6pQR4IAgcJgcMZFipoUGgQw/hux1AOGBTuJzB52UIxMM1GgsqZJFT2DarClKkZZpCCLYFHsw5mMdS0yWEfDr8ZZAqC+bFIw8+C7UIdhHgfbYdsMOt4vKiZkdOk/5+9twFws6zyxX/Y0ol2tCmtTmgrE2SgsaBNkW2Dlm340AZFmrpFRkAJu16Y3T9uB/zvMu5fZfB6l+zHlYBcHfyvy6ACg3ZtqChBoU0vVQICDXehDWuxaW1LqqVNdZAMDM79nTdvMpmZJJPMTGbycY4OnXk/no/f8zznPc855zknmnQZgpmriwy13QMnzZI9PTQN2Me+PuaK1Gvk8DOJppVi1ee+P6KdNA9TSkOgj87NlKTi4i/TNaa2sReKYCEPD9fBFvL/Bn5KikBNI2CDw94E7MvdEM6D3V6c0VjsTIwe5gnyng60032iMxxGh0OAoMk0Z9HayFtShjBDXpEIoYtuDml/sRTisSS8qRTsZDbWrBMZN1V5d4UslwJWr78PEa7nRJy+sfTBLYWigQCSvmBawGN9+dvBFnIjZ8llOMKD80pZdvgC7LePmzW6b/R0e2kCZkuEabKN3R0pnt6MmHy3lBbqM4pAbSPQoEIamWBgHSLXPIjDxvjNwxp/AL5CvDMZp+aMDFe4imjNekJIunmqiL4TQW/lJoDVQx82hrEYQ9wp+8NR7jJ58ot+ab5gGJ1paXHMoyMujJB8+McEJKFksB0uvxXdgU5082MjYTZKkdGKN4x38zLscd/SBxSBKkbARoHDjz73Ddhubghb1/XQ77TAZOdGLJ6wUIgztnBwdvQhmKSwEqCQ1uMu2k+LrR09YYbVGPFUCkH5u0B12UfjDLNBDbmTPLDLZ6f/bRdcPUWrS99M9qE76KBvmD37cP52kNVE+UMhLpcKsR+ri362UQ83wz64XWH0hgNwGy9aYLGIlp1vOsbr1Iiq9A9FoGYRaExzJ4fLzt1fbO82bL5/M7btjiPcWVh9ngiSWXRFcgaZAhuZRIrarqJk4+mpRFpzZjzH3aRos4QFj0s0acruM1sDd5E9dPaPc0sZDYYoNFrh6uhFqNuKQE9u2wqULFo0CpuJzO1UUhR9wzSCgaYKym8RatCcXQG0uxw8ucmelGLqlFomg0WBLullRaDaEbA4OxGO78a2zfdj8xN7EeMmx16o0RSWvNyYUZ7Jkt1hz+EzKSTITzIkv1tlDYrwQgFv+E4SYWq4QhT4yEa4RDNchHyOWqjh59IlJcN9iLFev9dprGkrV/84nM14MdbjR9xHbX+WoRVqB1tILZrws+HGk5Plq4QHHUIMdZQ0NsNh9DiC1CqaD9IVorvHh2gHw4fk8q7hUvU3RaDuEGhYIU1G0mqn74acKHQUF5ts9KWw02cje+IpGUZf2MpdXvH36AwCF0+RGqcvSYkQTyrRfOEoOI1oLsgwHz7njNLfLJ5+OBkKIBClyYB/RqlB6zE5uYVcWE6oGsTfLRS28vE+WHmqyhJG0Dw5lgj2IpwRzCw0PfDTEDHamYIIYqMZeboCYkZtYjzjMJYMoUdOaWUqLFZ/2VhkatR/FYEaR8DqgNtL31SXvbhSi/zIzfCyvdlj5AmEeKTbyRPgGYr2cXNp/BFHMJjkCW871z3Xtp0aJ/5tUCIIv5wQJ1tw8AR7lJp2Y4mSF3X5g2PC/FhE0IvFzDWfQF8gZJz8zstH0jVQ8guhu9eOrrQdNn21SDt4DB2OSJA8Rx5lTMhe1pG3hjh6u3hKPYeniA9ulij09nbE0dkxth/DD+lvikD9INCY5s5yx4+BH3u74zxyTtW+7FrJ/VzdfeCJ+OJk8ZBZ0lThdaGHfCZlddOHjFH4875lp5+ZhTtpB0LUkIW76OTLd33c3fbQpyRFQ0Y3j8nb+G67n0fkO3gIwEqhLuWCv89k4hYXvI5utPPwgpfvBjy5NTnQEWhnUF4P36NzLo/St5OxG7tZvtfVbUe724U+u50fBQqIBaQ0F7VoTgq2riAFO4Ym8ba76CsnHwe2bHT9uf0sC4u8AOlFRaC+EeD68ffRhYEhbnh2xuATNjf9snL8MFyuFLq5PhNcuFZ3D3rc8hT9uHq6EfFx3QaovU/xIE4PTZhyy0cTZoQHlFhOKkn+wrA/o9mWxdNt8BaPq4+aNPFb88JBgc9gDQUQT4Vl00rBjr6xvfKM8MiAt3A7aI71d4bQ6SH/I/9xsg4XMwOMEQTJJ7q7eslrPaC3G1L2boY6Yvmm/GlU1ckwJPS/8/W5eJCrQAP1siJQJwicIMdDK9mXjRs34o477gCPqmLRokWVrErLVgQUAUXAQOC8887Dli1bMH/+/LpBROKk+d1R+sEWEp3qpqvaEUWgJhA4dOgQFi9eDIbgANpkN5FDu34AHNktMTgm1ZeGNndOCjl9WRFQBBQBRUARUAQUgQoioEJaBcHVohUBRUARUAQUAUVAEZgoAuqTNlHk9D1FQBFQBKYRAVcglg6pMY11alWKgCIwswioJm1m8dfaFQFFQBFQBBQBRUARyIuACml5YdGLioAioAgoAoqAIqAIzCwCKqTNLP4NV/vu3ZM/7dJwoGmHFQFFYFoRiMcl/uOYACHT2gatTBEQBFRIK2UehDtg9/SNCQJZyqvV8owc3/cGxzKdUDtjFgXi6WaGfMzEEDabHEeAsdQyt6aiHy+++CKWLVuGhx56aCqK0zIUgTpCIIFeJuH1paO95vSL1xk7sZOxaUes1Qr13OATjF82LZRkLDdn14gMC9NSbwmVvP/978dXv/rVEp7URxSByiKgQlpl8a1A6YzW3dmJvsTUFO1m6oIQ8/VNBy1duhQvvPACLrnkkumoTutQBGoIARuzEtgQCcVGtpnp4EJxCVINFF+rU8sXagi4ijT1ueeew5e+9KWKlK2FKgLlIKBCWjloVcWzScQizG03RW2xWJk5gFlhpotEk3bCCSdMV3VajyJQMwjY3W4gHGbCp2FKMZVS1Mno/IxfW3ytTi1fqBnQKtTQU089FU1NTRUqXYtVBEpHoHGFNOax62ReO8l7aXd3oC82ARU/kwH3+txwuVz8cTNNScxIcxLz82//8I440eOGxdaBcHZcmCrK6UOID6fifUyB4oKbDNrt9qJzOEEoIn4PnE6WxR8vNV6pVBh+D9NFxaJMhsy8o5KfbzSxPJ/LyZx+LqbL68zmDc19LNHHtC/ePiNX33SaN8TPQwS0MD9ESopAQyBg8AgX7EztZmX+2q5QvHC3nR64EyGEc3ZgsWCErhbkH6PWarTLAQ9TPHUxr6f3631j+UKqD95cU2Kkk2s+mE7DxNzDfr4nvMXp4PvkI+Nxv3J4muQI7SRPM/iiu515hs0OSZskfVRPO3kak6TnIpGKoot8qyubtLMwTNNxx8rUVQEmqVdSBGYagQYV0mIUctpx+/YXcXzgOPZtvwuf8nYNJ/UtcVTiPT4ErH6EIxFEwn5YmVOzJy5JjV1IhiOmtosJy8NRDBzmMxm5LRZB1M7k6xYmM+4IwNIdNgSXMIWnWGeXIbwhQQbMnJy9Ub4XYdkBP5mam0xecoaSmTFpcrDTOaalYX8Xkp1hRNmmUDsZX++wsGg8zHI7/EA3GaVtzNuVvSCMT9KE2ZkfVEkRqH8EuPa7PLjmniex7/gAjr/4CP7J6+Umq1DPuRlzxRDK+qXx94iVeTfzq7ojvRG4e5km6nPt4/KF3BrDXdwgUmCLkrdEyVssgfHdJ8rhab0dwoNC5FvkXT3kYT5/1u8sRQGOnA2haA+Gk+ikEO3mJtbdi25RGVYBdXd3U4gdy1+roGnahAZDoDGFtHgIwecGRg71i2QeBZlnvlmRpPCVZFJ0VzoJMZMjt7viFNgoYXFH7IqHTaEvinDUgbVrEryX3lEmImFYKMhZDX8T7mQzjMlG7Zg9hpiouKjhsyTjiMsrkng52A0jl/I4ZLVYkIgnjJ2xrb0Xfe324TdSIhSSKXf3oH26JTS2IrM7VSFtnEHU23WCQAx9wX0j+zLwHPoKatOYGN3rRDQUTb+TCHPDRg17zhLOLcxJYcgzgXXs8lOw6zALtTpZfpIJ28eBvFSeRnEsHCMvzAiWDi+8oLAWT5dvsXnR1eFErtiZinSjI+xFT7ezYEL3cVo35bc76fcr1g0lRWCmEWhMIU38sOaNgr7JDkdZDE8YG02lOe9Y7VYkk+R2FKo8DmFWrCPO3aq1Hd0dLsSoXUvxf1EKdy6D86aQpHmjy+2BxyM/1MzFWK5IWFbuuLuBLpokOvriNJXQNFvCbHF29cAdopnUG0A4xT7ZM5JdColeH/yWbvTQQXkmSLC55pprIGZPJUWg/hGwcf2N9muaR01y4fVnJS+w0Q9NWEcyHELC5UUhfY5seiZCFppeE6EeCks+tHuF55RQSqk8LZVEkhtMVmGSFTZbMr3ZlCts84hWc6Pa3RGGt7cLzhI2oSW0dEoeESFN3TKmBEotZJIINKiQRgEosA4tWfDmYY0/AF9h3lkmzFa4PVaq+xNI0qSQ5I7MReZrpwYtRt+LUMzJo+fpIi02+myQGYdC8sP7CZpiHXLPAoevjwJdF6y9FLp8QcOHbFyiNs5P82qvJ4pO+oQEoiLxmUQGmUqyTeMWUpkHREjbtm2bCmmVgVdLrToEbPDRTWFNzoawdV0PXS2KSCNcvx5LBGHu1OSkp2jWppZSCHdS09ULeDq7EejtQWdJ0lHpPK289lpgsaQQN3am1UMioEWj0eppkLakYRFoTCGNw22n0BPbuw2b79+MbbtppqR/VxHWmWeCcIdo4Q4xR3JK0jaZ2d3aaM5McaFHwnG4RCIT5msl86WfhuGPZpRIBkUT5HARSYTprBqSCzSX9lHIs9i9NHX2wRMJIHumIE9r0peSiAZDFAStcHX0ItRN59eeiPm0BTZvDwI2CqMlbZ0LVjLhG2LmFC2amhEmDKG+WGMIWJzUyMR3Y9vm+7H5ib2IBdthL9oHB9wuhtPghi0U4eGAyfhojQjGKjp8IZo6QzZ09nTA7Uif7C41aGtJPE20aKJNy8pctAwkyCsLKf0sTvrH+hDt4EGCmdo95hkPEdBEm6akCMw0Ag0rpAnwVjt9wNq9ZFaFOEix4WEQWGrLwkHzZBTV9n0RxjrKMFX6YjhjvQhE7GS0Ug4ZrjuJoD8ECy8YAiEZlNseRm/Q5E6JIPxyYlOaw9+7u2nuMJ6T3WauCYFm0gIMLdrTydNU6XZb+J6cXs0Sf/dQCLSTKdItZUYoGAymTcIzUrtWqgjMAAJWCl48qOR12UvaCDqpPYsF6Fxv42nPslhTDl+wUACjQMZ9Holaub6wuRkUgSmOqHmaPRXpQS/9aIeFqiL4lMLTaJx1O8gL6dJhUCzIpPA8vW4vUi4F2d6OODo7yBuKPDadt0STpm4Z04m41lUIgYYW0gqBku96KtIFN0/7yIkfp5OnLKmgclBb1ZlIO5i63XKiqRcZf1yeHqBfGjVnFjnFmS7RSeEsFk5QkMvYVRlhvKcbFr+b4TJoEmVYDEePn+YOPu/qgt/eCy991dyuDsQ6uuE1GLadhxUs6GEUcrc/oyXLtNiKdp4wjXdSa8e6PD12+LucI7tj9dDU6+CJK0b6nmYLg+xO169fr74e+SaYXlMETAQs9ENzxJ/jZs5dxgns0XzBha5uO7MYUECir2sfN6Rpq6aDp7u91FzJdS/a+xx08qcYVxIvKIWn0cRLHmYJkG8JT+uIor2H/mbjjK6jswcd5KW+qYrSPU59490+//zz0dvbO95jel8RqDgCJwyRKlmLhFy44447cPDgQSxatKiSVWnZNYCACGp6tL0GBqrGm3jeeedhy5YtmD9/fo33RJs/EwiIFk1cVyZ6OGMm2qx1Tj8Chw4dwuLFi4FFK4E2z8gG7PoBcGTyuapnT3+3tMZGRkAFtEYefe27IlAbCGiYoNoYp0ZopZo7G2GUq6SPokXTjANVMhjaDEVAESiIgGjQJKCtkiIw0wioJm2mR6CB6pfd6c0336wZBxpozLWrikAtIiACmp5Cr8WRq782q5BWf2NatT3S3WnVDo02TBFQBHIQ0PAbOh2qBQE1d1bLSDRAOySYrZya0qPtDTDY2kVFoIYR8DLHqoQLUlIEZhoBFdJmegQaqH4R0uQwsQppDTTo2lVFoEYRMFL8KSkCM4yACmnjDUAqyGTkXvSNXq9y3e5jAuQIOulr1Tk6ZNl45ZZ1P46Ay41AvKyXJv5wiFHRmU9vqkl80iRIpPp6TDWyWp4ioAhMJQKiRfP5fFNZpJalCEwIARXSxoPNwsCwTubbZFTuERQJIeqUQLVMa8JUT92FojVSiPP7eoyEyUpgIN2AZhzQiaAIKAJVjYAEstXcnVU9RA3TOBXSxh1qSSxsY7Lj6Igno6EIbIwIbmWiF6vNxvRLBQpKxRCOxs28eQWeaZDLwvRuuOEGzTjQIOOt3VQEahUBOTigPmm1Onr11e7GFdLiQXS6HUZuS7u7A31mLrt8w2tn+hQLtWXx7E0mZA9bmHpJ0jvlmiL5O1OhdPX5mafPDf8DAXi9fkRiffC5PegKUxsX8sHhCw+XFGB6lh4jwR5S0R60uxxwupxwUEvXlcl/l3169C8J9DI91LCpNYWgl3k+vcFhoTDmZ4L3XiMnXiLcxXa54JIftitbvLSpnTn8Opg6xtc3Mn9evIfpsGjWHW3uHd2UEv6WQLbHjh0jJt4SntZHFAFFQBGYGQRkQ6lx0mYGe611JAINKqTF4GfC49u3v4jjA8exb/td+JS3C6MtmlmoHG64kyEw7WaakmH6otEMas8znSTRetCOAH2vui6X3VgXXI529IZD8LsLqduknBi6aRa1Mzt6NMKfHgeCzGdX3EzKhO5M8h6JxM2GiFl2AANiijWvJNgOyQFoTYXQxTx63mCEz7ON3jA6OkNZYS4Z6kPCF0S0t93I756mGAJsk4MmSlajpAgoAopAQyCg6aAaYphropONKaTFQwg+NzBygF4MUptWaMyYWNgVy/qlpShwxV2eAkmDmbC4i473xeSxvNXY0RUK07ct/aKFSdwdyfhIrVae9+xMoJwMR9LPxSOI2tZirSWKtNyWQoTqMpfbThkwhIi9HV5R/pHsHrYxGs4KgRZ3JzpduZJYCrFAB3odAWoHp0ZCk92p5FJUM0KegdRLioAiUDUIyCEn1aRVzXA0dEMaU0iz2mGbN2rcm+xwmALM2BlhocnQCfFDo1ESkSCThHtcYx+TKzSf2iYk09C3zZpEpMePTh+FKW83Sjow6uDhBRHAaElNRsJIubtoxk1QoBT7ZBThOAVMB39NJJES37lMqyV5cDKRFQIt/HuEXBkNwNfrRE/APfJ6/l6XdFXMnbfddpue7iwJLX1IEVAEZgoBOeCkbhkzhb7Wm4tAgwppXp4yXIeWLBLzsMZPoaSgkAZY3V7YxYzIgwDBiB2eoqbLiUyyKM2hHvhTTvi6A+jtY1qSUrRxFheFMApj9KmLhuOG8Oj2Oqk4iyAlmjW7nECdQHtEaEslRLabUhKHXDUlTCmkWpgioAhMMQISfkM2lUqKwEwj0JhCGlG30/8qtncbNt+/Gdt28yBAp7O4xshGHzRLhAcGqLWyeTBhC6AhMFHtZVIq8yu1YYzMg0Anzah2OS2aQvZe0Vkip0/FL42+ZlEKj5TILBQoHTRlRsQfjeZQo0qq9yyJYc0Z42Agac3RrI2uw05TZ1eSfmz0VRt9b4J/SxBbyTggsdKUFAFFQBGoVgREiyZhOJQUgZlGoGGFNAHeanfD2+6F21GKfdJOM10Svd19SHk8sJc6chYKTZS2smKZzQELfbNi8r5o5Wg6NUgEpkQMcUNzRX+w3h7IYdBhcS79WL7/2qiBS/FEachqCo9WCpTWMPyMwOt2m+pBMYvyRGvIlLgSPCgQF7+3fAWa1+y+XnTzf77eeJGnSr8lGjTJOCD+HkqKgCKgCFQzAqrxr+bRaZy2NbSQVu4wiykx8aQIPsVEm1Gl0hzpdYTQ7nChM0SRy+GjWZN/U0DytAdgYaiOtKarHf5Oaq7cvM6Tp/SIoOaOfmSlNJKnT13xMJJ8Ny2SUaPmSdEE6qQAahZgoSk14OCpTppDqV3zBulv5vcU1x6ytPaebli6fegxpMrJkTA90aKpkDY5HPVtRUARqCwCcrhJfdIqi7GWXhoCJ1CzMVTaoxN7auPGjbjjjjtw8OBBLFq0aGKF6Ft1g4CcmBJ/DxXU6mZIq7Ij5513HrZs2WKcJlZSBMpFQA4OSPo69UsrF7nGev7QoUNYvHgxsGgl0OYZ2fldPwCO7DasR5Mh1aRNBj19tywEJATHLbfcoulWykJNH1YEFIHpRkA2kxoqaLpR1/ryITA730W9pghUAgHZlU52V1GJdmmZioAioAjkIpDkwSolRaAaEFBNWjWMQgO1QZMWN9Bga1cVgRpFQE6iq6BWo4NXZ81WIa3OBrSauyMC2ooVK9SMUM2DpG1TBBQBnHrqqYylGVAkFIEZR0DNnTM+BI3TADF3bt68WTMONM6Qa08VgZpEYNu2bXq4qSZHrv4arUJa/Y1pVfdIj7VX9fBo4xQBRYAIyMlOJUWgGhBQc2c1jEKDtEH8PMTcqRkHGmTAtZuKQI0iIFp/zThQo4NXZ81WIa3OBrSauyPBbJcvX65mhGoeJG2bIqAIGPHRNJajToRqQEDNndUwCg3SBhHSdHfaIIOt3VQEahgB5VM1PHh11nTVpNXZgFZzd+RIuwSzFbOnkiKgCCgC1YqAnOxUt4xqHZ3GapcKaY013jPaWxHSJJK3CmkzOgxauSKgCIyDgPApjek4Dkh6e1oQUHPntMCslQgC4uOhGQd0LigCikC1I6CBbKt9hBqnfapJa5yxroqeiglBGWBVDIU2QhFQBAogIFo01fgXAEcvTysCKqRNK9zFK3vjjTeKP1Djd0VAO//889WMUOPjqM1XBOodAQkVpIcHRo7y4OBgvQ97VfZPzZ1VMiy//e1vsf1/78BlGz5RJS2a+mZIgEiJ5C3H25UUAUVAEahWBHbu3Ak5ja40jMCzO6N429vehrPOXKawTCMCKqRNI9iFqjp48BC+e+/9sNlaCj1SN9c1knfdDKV2RBGoWwR0Izl2aP/0pz/hP34YxMDAAD5w9oqxD+iViiCg5s6KwFp6of/1X7/CPd+915j49U7i4zF//nw92l7vA639UwRqHAHRommC9fyD+NCPH8bjO36e/6ZenXIEVEibckhLL/D5F3ah7/ubUO++aBlEhPFdffXVGsm79CmiTyoCisAMICAhOFSbVhj4rdu240cP/URP6xeGaMruqLlzyqAsr6Bnnt0J2ZE0EunutJFGW/uqCNQuAp2dnbXb+GlqufioiQXoE+vX4S1vUX1PpWBXZCuFbJFyf/FEpOEENIFDQm9cc801erS9yNzQW4qAIjDzCIiQphkHxh+HF3btNqxBevJzfKwm+oQKaRNFbgLvSSDXn/7sMfzs0a0TeLv2XxEhTU53avyh2h9L7YEiUM8IiICmGQdKG+Ff/WoP7r2vD6lUqrQX9KmyEFBzZ1lwTfxhORnzw80PQnYejUqScUAFtEYdfe23IlA7CKiAVt5Yxfftx92938VnPn0F5s6dW97L+nRRBFSTVhSeqbkpqmBRCTeygJZBMhgMasaBqZlWWooioAhUCAHRpOmGsjxwf/u73+Hbd38Hx4//vrwX9emiCKiQVhSeyd98/fXXDVWwqIQbnWR3un79evX1aPSJoP1XBKocAcmMohkHyh+kY8eO4d9778HRo0fLf1nfyIuAmjvzwjI1F1999VXc1/d9HDr08tQUWOOlyJF2ieStR9trfCC1+YpAnSOwd+9ezTgwwTH+/e//YGjUrmj/JBYvXjTBUvS1DAKqSavQXBCVr0xUFdBGAqwCWoUmnBarCCgCU4aA+M9qWqiJw/nHP/7RCNK+N75v4oXomwYCKqRVYCKIqldUvqL6VRpGQMydJ5xwgprRlZ5FAAAgAElEQVQ7dVJUHIGPfvSjaGpqqng9WkF9IiACmgS0VZo4AhKkXVx9JKuO0sQRUCFt4tjlffMwE6WLBk1UvkojEZDd6c0336wZB3RiVByBL3zhC0YyaCVFYCIIiICmeYYngtzId958803j0Jxk11GaGALqkzYx3PK+JardBzghGyEPZ14Axrmou9NxANLbioAiUBUIaMaBqRsGiQ8qidl///vf44Pnuqau4AYpSTVpUzTQotIV1a4KaIUBlWC2cmpKj7YXxkjvKAKKwMwj4PV6IeGClKYOAQniLsHclcpDQIW08vDK+3QmUbqodpUKIyBCmuyqVEgrjJHeUQQUgepAQPiV0tQi8ETkSU3MXiakau4sE7DRjzdiovTRGJT6t/ikaT68UtHS5xQBRWCmEFAtWuWQ18Ts5WGrmrTy8Brx9OM7ft6QidInARkCgYBmHJgMgPquIqAIVBwBCWSrGv/KwZxJzC7B3pWKI6BCWnF88t4Vk92PHvoJtm7bnve+XsyPgITguOGGG1Sblh8evaoIKAJVgoAcHNCMA5UdDMnCI7HUNDF7cZzV3FkcnzF3NVH6GEhKviCBbCV2nAaJLBkyfVARUARmAAHZUIp7hlJlEZBg75qYvTjGqkkrjs+Iu6Ka1UTpZQCmjyoCioAiUIMI6EZy+gZNE7MXx1qFtOL4ZO+KSlZUs5oovUTA8jwmu9P58+fr0fY82OglRUARqB4ERIumGQembzwyidklGLzSSATU3FnCjJBE6d/57n0QiV+pPAQeeughPP3009mX1q5di0gkAhHYhNra2nDVVVeVV6g+rQgoAorAFCLw/PPPY9OmTdkSJdtAIpHICmonnngi/uEf/sFIa6dUGQQkS4+YPj995ac0MXsOxCqkjTPfJFG6aNA0D+c4QBW4PTg4iFtuuWXE3UceeST792233VbgTb2sCCgCisD0INDS0oJ//Md/hOSbzEfr1q1TAS0fMFN8TYLBy/f2U+2fxKn21ikuvTaLU3NnkXET1asmSi8CUAm3Pvaxj+Gd73xn3idlV3rllVfmvacXFQFFQBGYLgSER11wwQUFq/P5fAXv6Y2pRUATs4/EU4W0AvPr4MFDhupVE6UXAKjEy2ImuOKKK/I+/ed//ucFBbi8L+hFRUARUAQqhEB7e3vekkWAk82m0vQhkEnMLsHiG53U3JlnBkii9Pv7vl9Q9Z3nlSm5tG/fftzy3/9xSsqqpkIGXs+fLmv+SQvrsr/VhL22RRFQBEpD4HWa2ubMmYPRAVZPO+10/KP/X0orpK6fGmLvps8nT+KRPvTjh4182I2cmF01aaMWVSZReiHfhLpegxXqnO3kk2Gz2UaULsxw2bIzK1SjFqsIKAKKQHkIzGlqysuTnCvOLq8gfXpKEWj0xOwqpOVMJ1GtShw0TZQ+pWvMKGw0oxMBTZiikiKgCCgC1YLAaD4lm0vZZCrNLAKNnJhdhTRz7v3iiYihWhUVq9LUI/C+9zsxa9asbMGjmeHU16glKgKKgCJQHgKt9lONWI4ZUj5VHn6VfFoSs//HD4OQiAGNRCqkcbR/+rPHICpVpcohMHfuXJx++hlGBcIEhRkqKQKKgCJQTQjIifPlzhVGk2RTKZtLpepBQBKz33tf3xi/wepp4dS3pKGFtEyidFGlKlUegbPe936jEmGCGhSy8nhrDYqAIlA+Ahn+JJtK2VwqVRcCcR6wa6TE7A0rpInKVFSnokJVmh4EzjhjKZroh5bZqU5PrVqLIqAIKAKlIzB//klobbUjs6ks/U19croQyCRml2xA9U4NKaTJEWtRmYrqVGn6EJCDAhd/9BKaO0+avkq1JkVAEVAEykRg5SoXZFOpVL0IZBKzHz16tHobOQUtazghLZMoXVSmStOPgDriTj/mWqMioAiUh8CyM8/S0+flQTYjT0u6xm/f/R3Uc2L2hhLSRDUqWQREVaqkCCgCioAioAgoArWNwB//+Efjuy5ZguqRGkZIE5WoSNyiIlVSBBQBRUARUAQUgfpAIJOYXbIF1Rs1hJAmqlAR0EQ1qqQIKAKKgCKgCCgC9YVAJjH78y/sqquO1X3uTlGBfvfe+438XxWnuStw9Y0Xw86KBmOb4f/BbuTPWlnxlrCCU/CxG67COc38Nfks/u3rIRw0q5238mp0rl2c/mvwRWz62n/gBROeBR/8K1x/YQvvHcGOb30Ljx2ejrZqHYqAIqAITD8CLRdci44PLWTF/Xj6O3fgx6YipvWjfwvfB4R5DvPB4Wf34aGv3YtnSjxYOMxTDyJ0xz148vh4/ZyFpZd9Hu0O+TwfxqPf+DZ+/or5TstF2HjtSliNP1OIfudreDCjPFrswef/8mxIqw888r/w7afGrWi8htTcfckWJFEb5Hv/gbPT8e5qrhOjGlzXmjRRfUo8lWkR0KpuJhzGgYQZmdm6AAuz4X5mwdYqQphJs1uwRHiUQbOw0LYg/WvqtzhwJHNd/1UEFAFFQBGYHgTeRGJfRipbANvC4Uwtcxe3mAKatMQC22KTX/OvuS0LDAFNBM7EwcYT0HLHRrIHSRaheqC61aSJyjP44I8aOA/nAA4cTgJtIoEtMASx54yd3wLYbbnDboWthRLcQblpxcIF5r0jB5GYOTVgPawt7YMioAjUEQKHt34bX90uHXqT35XKduz44YMUtVoodM2GbYHozURo4wZ7Sc4Gm1cWUkibxXvSnOwGe/Ag4rrBNrII9fe/io98+MLKDlaFS69LTZokSheVZ6MnSk/u+y0V4kLNFL5MVdrcxbCldeXZqWWzt3Chk2alhTn5NUkm0dh7sSw8+osioAgoAkRAhLPKC2gG1AlukjOGEEMQE+KGerFlxEjMtp0Cm3FlLnl8Wo+GI4eQmAbvnhENqdI/MonZ//SnP1VpC8dvVt1p0kTFWdV5OGfNw9I158N12iKqsS0YTL6CA/v+D3Y8thMHubCG/cUO0u/hnrTfQ9N7cfmN62G4KCSfQs/XH6WnAqmVPgifER8E+k18g/5jGQ25Oe5vJvYhgWWGj9zCJVSLP/sqZtkopMn9wX2IxVvgaLPAskBU6L/GKzSLWo0ZMYhEPF1YU8sKeD7yZ2ijZGeZnUK/tPeZxxF6aj8yLhmzFrwPl1x6Ps6ihi55YBd2vDQPl1x4GveAub4bs9Dy/vPhXt4KG02qzRQfj8izP92GF16p8LbUxEP/UQQUAUVgogjk90lrQusFXlzygVY0D76C+M+fRf8HLsY5tFz0P/M9/M+f5MbjnA3raRfh6lXLsISMNhl/Ag/98AnsyydQDVDQMg0hsxcuwkL8Fw43LYLd2GCnsCd2GHZHK2ZbW2AzDCFiFk33LMUNdpp7z8NpF1yEi85sxcLm2RhMJZFM/Jo893Hy3EylpbW/afEKXPTB98PO74iVcmL/kX14/n8/ivCLx2fQ77q0kZSsQn/o78cnN3wCs2fXnshTV5q06k+UPguta69E+4coOBkCzSsYtC6Gg4v6qk+8D/M4544fPAiuTdK87KKjExmWZOYWF+USUyk2r0XU4aT+/dgzSkAzinj1sLHQhZoXLDDKt7YuoicD6civEXnRPBWwkLuxJirSDGYgREHsMBdx0xm45IqL4bQvxOwjuxB5hs9bW3HW2qtw1QczvhDzcM6lfGYJFfOpw4gftOCcD5F5GOUMU9PSj+OqdSvhsPM9LvBEiv4UbSux4TMfx1LWraQIKAKKwMwikPH9GvYBG689s077MDZ86DRwv43+fS8hufRcOEdZKobLaIFz1QIKS6CwYMHCtvPhXbOoQBVJxA+n7SAQn2JpkmywhbEOHkbsqYPcmgvxeyC77rl8xlSkJeJpvr7ggxv4rVkKW3MK8WeeQizZnOa5f3lxlueW1P55K9DO78A5jsVoFh5PFV+zbSlWf/JKXLS4dKwKdHRaLv/qV3tqNjF7XQhposr80UM/QdUnSp91Gs45M72C+58J4t++ew96H0ufubS0nQ2HSFGJ/ThgrM1hE+WCxVwc2aksizK9MDI+CIMHfo0Dead6ekEZtxYu5kJvwpIWs34KgwcOH04LhLPpp0YuYKUgZwhXXIiUH/mOCG39tOsfQXTrQ3jspw9hRyJdkY3CniFbNZ0ChylBJp4M4cdbf0RNmqmnTz8qD8Fx9hnpPhx5HN+7+wF8+77HqeUjNZ+Bc05TKS0Llf6iCCgCM4BAM875zE24+Uv/wJ+bzJOd4zfDtvQ9ab42+F949MEwHvnJThwZvUPNFnMET//wAQoLP8Mek0VaLCPNl8M18vCAKWyB/Fm0ZPNazO9Akj5nB/bjiPGd4GaXBwZmLWwxN9hH0hts8lwbd94papCSex5H8KdhPLj1v8jN5RVq5EytWyntn3fm2bAbzdyHR79zL+79ziZEjc2/FU5qEGtDTANqNTF7wekkQ1BLVBMnOJvnGapioSOJtLPn8cOvUHm9mEuNmjORn45zAVJ6OcuecQQdxJJW0Vr1Y8+eV9HWRiGtlQ9SEFrYki7swEsHC6ic5ZQQd1XcAdGmiYXUwi1Mb8UMBvBmwkK/Bx7nni2HB3hoYLG5BTTMpKSDYdx1Wxiz5r0L9sXvw/Kzm83FyntkLlL7QHOzKUAO0hQqLOBNJI1/c5mPaAXTU23wyCtpwZBm0yQZlY3qZ8MUu6s+o0ULjEqKgCJQjwjMQvNck8+JMCReG0maFA2+lq+/A+gXwYrhIfpFSMv7zPB7rx6UTXQrRSHy5wXzzO8A99AH97Mu2YCb34nF74I1ZZ7szJ7KH8ALP/wGXuDGfMGSVtjfvwIW8tl0leTdRrNLa//ChaI9IPW/giOGj8vxtIWGnwsLN/bNdJWpFf/lwTcGa85XfZxpkh6bav/vW97yFvzFJ7x4xzveUd3aNMvcrOgyOGhupfhv+jcLqAEnvYoDB6nI5lZHTJRzZ82mHwCHKfVrPP2LFFXk9B9roRZr1oAZOuMw9uwrHLDnOLVl/RQCm+WE52nvMVXi9J0Qc6YsdFbloLp8of0MVpmeDkfIHNIeC3Ox9NIrsWE5zZ2FJsHsYS3Y4JvpngymRvuYsW9mAbMdG/DFL40sbLalYOmFatXrikDFEDjG7CSJxMtoamrCW9/2Nm5gbDiBPEapnhEoFCeteJ+zLk4Zfj6Vxz6P7Dc30dzICn82N9gH4rIpH0BCTu/TSW023WEcA6b7ibiSmOx31uJzcdUV5w9vrPN0Zfz2z8LsJpM/N58N35fOHlmK+U2rBSHN3noKLv/kBgqouQqEPKBU2aW6+TqecMIJxlHb5ua51XtwYHDYQzQLfO4ImHLbkX2HkGKARYuosLmLWcI5NbjnJaq436QpdCUd9OmbsPA4DJmqn2bLfP5omYkmWrHBs9FGCXDJB0yTY/Yd7ogOU+tla4aFmjKHYVNN8WBRusBZrefjEkNAo0/DY5vw0HPHYf/E/4NL7IVmsckdLEUU4PRtCz2TCaubLmeQ2kQlRWAmEJDNUuLlQ/jNb/bjN/v5w3//8Ic/wLv+L7j23olv3fUNyCZw8ZIlePe7T+EP/UPf/e6aY/QzgW1D1UnebrDvIqyvbDxMbZljibgCrzAPDdBXTU6YkRIHDmNwFYU063vgNO0ZSW7K0wLTPDg/Ygpoyf+DTd9/HPHmc/HZK86mXi4PldL+1EE8vX2X6QtnlsFviWH5rHI6/fS2mj04UDdCWmaOfPBcl7EDlmB2VUf8IJiuoNSaGcZCqrypZTIamkLKFNLePLAPBwbfj7bmBWg7Uyz/XJAv7scAZaC0insx2k7jkWt5q6A/mtn7gd8a2rI2asusC9OOCIMsn5pyg4xTnMspnVkXmouXWjZzK9acDY54ENFn9uOVgbnGSdERlBU8Z3PHJVq1AX7ARk8r9k06zmoGky8i+tRuDDDcx2lnLjJYSyrT8dFl69+KwBQj0E8BLFcge5kCWlarbdbldK6Ac0VaY/CxSy5FcPN/4NcvvWT8CMmG8J3vfGdaaDul1fh3gbm2pri5WlyVI5BRoCFjJaFlYQz7m3AfzE00D2Vxx5D2OUsdym7K3zx8iAITDwaQcy80JK/hU/mZ2JhyNfnis3jhMEU3fnJGc+bx2/8mBgcyttkjPLDwS7xEf7eWpWcYh83EdGt+tibcy0q/eOay9+IT69cZm61apNFjVot9GNNmSQchgtoPNz+IoaGhMfen48Jsi/gRLMgKZRikzwJt+kkKK6IZszIMxSzueUQQSitfTTu/NI7Hr9OCFU8DLafww4AbMdOkmTaFLsBZH5hrLLjC/miZXiapLWOltmEVr2jqMjq9AR7XPkK/B9OPlCua8Xkyuuvs6mNNUtnAPJMZ5CDYn/bBEK3ekjOXYcFLh+kCN3qvdhxHxFGDD81eyBAcs3Zjn20FLuFpT3ky/uPb8aKYX5UUgQoicP9938OLsd1Fa1i8eAlEMMuQCGsiyD0ZeSJ7TXjKb5kPWH6eeeZp4/rb3/52dPzN5zB3bja1R9F69GY9IED/21dEj0TuSW3WWa1z0T+PwssUdk18hweZnirzoc7dYOMVnt4fwdrNU/lG/cOiU8ak2TRv3ghPYcN/uIT2H5EPwnJ+PywM90Gr6kvJxVh96cdxllh4XtiE51+qXkvI2SucuORjFxsbq1qluhTSZDDOosAw58QT8f1NP5wZR0H7+fjs35w/PC8S23Dn//8Unn4hSQHLiubll+Czc1/BbAaSFUrteRaxrGFfFlvaDEm/fENwMsOW4chLaVOo1So3ivujpSvn4QGqxcH4ZGnqp89bjgeBBD5My0/G7UEGUaTizSCJep2kP5uVP6vXuhkHiOFCZpu7qszMGaCv3Iv91PjRZNp2Ma6/MUXfWVGbDQuFol2LPfsSUm1LYbGejQ1/uQDJZimX1M8YcS8U9qkzm6L/KAKTRuDccz9YVEgTX5XLPtmOE+fMGVHXR9ZejAO/+Q0OHjxQsA0fOOfPVEAriE793jj83H8isep8HhRYCNdnNsLFTascm5oqGjjMU5w4LSv45W6whf8foEnEaTdry5zKN/6kReTAIJxts9F85vn4WPIw/20xRDfZ9mfYdyntP05NXPxC5qSebYP7M1ejrX+e6efG06pPvlTg0NpUITDxcs51rar5bAPS+9rU/5U4bmeccTquvIJMl8JaddCb2PfIA9j05ItGrBmrnac6KYDtee5n+N4P/zMbHFbamtjDHZTZ6NTB/8qaJ99MvERTqHmD/gAZ4a1Y/17Nxl7jU6NThtDvQRZ6ho4c/G1Wy4aDj2PTYy/hCLlO8+LTsISCZvA50wPBMs/0ghjAiw/eg+/9+BeIPPkLhL7/PWx6Zqwb6cCLD6H3QQqikhCUseFskl8u9gv0/fvDeEmVaMWGT+9NEQL2U+m7Q1NmIRI/NOv8+WNuz5o1yxDeCjkcL6QpavV5a8a8pxcaAIFXnkDvv2/Go08+i8hjP+LvjyI+rMSaPAAUrkRblibJyZmrteIhMzk8YNJg5lS+8fereO4nP0L0QJLWHIZYOm02nt+yzfx2zKZ11jzwVUr7j+9E33d+huge2bQzugB31xKIXMJxPHJw9CGxbHNm9JcPX3RBXQhoAuIJVN1X1B64ceNG3HHHHdyFHsSiRYUC91V2PA8ePIT7+r6PP/7xj5WtqCFLn4fW959iaMX6KUy+xEjWCy64Ftfz4IPE1Ql+7V4zZ2hDgqOdrjIEfvbTR/DM07+kH+RrI1rmcn0Qno9+rGhrY7t3oe/+e8c843SeDff5F+QV8MY8rBfqCoGmljPgaKHAM3gce3YxC8usM3D5328wssP0P9PLjAPVHVqovPaLqFDdZkMxa37sox6Iy9N00KFDh7CYcUyxaCUdvz0jq9z1A0qzuyftclXXmrQMYosXL8JfXfMZhuh4+3SMW4PVMQ9nXfhxeNd9HBs+cgbmNfGwQ8YnjbvAdFydBoNEu1uVCCSPHcPPd/xvLFu2bET7xA/tw2tHMdg8PXC8dxlEmMulZcvORDRKc1B8b5439FK9IzB78Z/Rt5b87y88WL24CU3295gx0ujEf2CsRaHa8Kj19ufiKRrvdobYmC4BbbrGsiGENAHzpJNOwl/6rsb8POaM6QK7PuvZj/BPdhnHsC1tH0fn318Hj53byEFmKfjpExgZbKM+EdBeVTcCov2K7nzW0HR1f+V/4FLvJ2A/9VSj0Rk/NGHwpZAIc6fSbCr0NsZQu+RSr1GmHDBIvPwyeu/+N4gwqNQYCLxKV5WwkSKGPml/+Xl0mSEu+uPb8GgN+NrWevszs0xcmsS1SVyc6o3q9uBAvoGaN+8dhkbtO9+9D7/93e/yPaLXJoDAqy8GcfvXfoHTWhlol+fPB1M8qbrv1zis5wEmgKa+MuUI5PHouOTj63DXN/+XEQ8tnx9aoTaIMPcXl12Onm98HRd9eK0hqI2gijqPFGqVXp8xBN78LX5+9+14nj679gU8cc+DVf109t/Dw1nV6a01Cqlabz+7I5EcPn3lp2h2nBl3qkrPvYbwSRsNYopBu7577/04dOjl0bf0b0VAEagDBPoYbkOo/YqrCvbmtwz8+a6W9Onqgg8VuDHeu5Enfo7wtq3o+odR6TUKlKeXFYHqR6D6fNJkk/SZT1+Blne9a0bgU5+0CsEuJo6rP30lJE2EkiKgCNQHAqnXXsuaGsV/LBOQtlDvJiqgSXnjvWunb5KLIT+EcttVqC16XRFQBMpDQHzMxTI2UwJaea2d+NMN45M2GqI5jIUkNmyJRqykCCgCtY+AaK56vnmn0RER0ERQmymynXwyT3xeaFQvGQv67k9r9maqPVqvIlBPCLyLGT/Ex1x8zeudGsonbfRgzmYoZknMLjbtZ3dGR9/WvxUBRaDKERAtVTS609BayY/jvdW36fJc/DEGeE4fJpAwHlbrfIgQp6QIKALlI7Bo0cmGD1qhuIXll1jdbzSsJi0zLBJX5eOXfBQSnVhJEVAEagsBCX8RevjHhklRDgBIwNpqo9x2hbc9Rl+1x6qtidoeRaAmEBAXJXFVahQBTQaloTVpubPyIx++EM3Nc/GzR7fWxGTVRioCjYqAhLgIh7fC7b6A2rMPMYvA2bC89a01AYfvms9m2ypmUMmAUI2CZU2AqY1sKAROP70Nn9zwCZ6gbSyxpeE1abmz/IPnuoxkrEqKgCJQvQjI6ez43l8za0A6X06tCGiCaG5bpQ/J5HBan+pFXFumCMwsAuI7LoFqG01AE9RVSBs19yRasfipiRlUSRFQBKoHgcDX/tUISiv+XJ03/l3N+3VJH+SAg2gGpW8SDFdJEVAERiJw9gqn8U1+y1saU1xpLL1hibP/rDOXYQ4jGG/6YRBvvPFGiW9N/rEWWws8H7lo8gVpCYpAnSDw+9//nunc3mH05j//z05s2LCBqZlcddK7dDcOHDiA3/wmjiuvbMeSJUuQ2+e66qh2pqYR2LU7hl/+8plp7cOF57uxevXIVGzT2oAqqEyFtAKDIOklxEFRgt4ODAwUeGpqL1t4ytTe2jq1hWppikANI+B0OmG32xEMBrHpB0xYXIcka371hz5k9CwQCOCGG27AMWrXrFZrHfZWu1SrCCQSh6e16eJ6VG95OCcCoAppRVCTNBPX+D5tpJH64x//WORJvaUIKAJThUA8HmfC8jgPBrjR3d1tCGmNQj6fz+iqCGi5ODRK/7WfioC4Gn1i/TqIRUtJfdLGnQMSzViiGkt0YyVFQBGoPAKdnZ2QHyGv18sTkM7KV1olNYhwlum7aNWk/0qKQKMgILlx5YCACmjDI96YnnhlzniJaizRjeczDpOSIqAITD0C0WgUt9xyi1Fwb2+vYd5sdBItouCimDT6TGiM/p9IP3DJAiSuRkrDCKiQVuJsmDfvHYZGTdJRKCkCisDUIhAOh3HbbbcZhYo2qZFMnIWQzMVBhFYRXpUUgXpEQLL+iA/4qXb1yR49viqkjUakyN9z5841fNQ0MXsRkPSWIlAiAqIlOv/88w3fKzHxacywwsCJkJbRLq5fv14FtsJQ6Z0aQ0BcieS7Kj7gSmMRUCFtLCZFr0g6ClHJSvRjJUVAEZgcAkNDQ5MroAHfVswacNDrtMviQiSuROL7rZQfARXS8uNS9KpEPRbnRomCrKQIKAKlIyBaMzm9JaY7ORAgZk41bZaOnzwpGjU5BZqLZXkl6NOKwMwjIK5D4kIkrkRKhRFQIa0wNkXvSPRjiYIs0ZCVFAFFoDgCYtoUc6YIZDfffLMRXkNpcgiMxnL79u2TK1DfVgSmCYFFi042TJziQqRUHAEV0orjU/SuaAQ+fslHca5rVdHn9KYi0OgIiO+ZnFYUarTYZ5Uc+wyWol0TwTdzGrSSdWrZisBkEBCfbjkkIK5DSuMjoMFsx8do3Cc+8uEL0dw8Fz97dOu4z+oDikCjICCmTNGeSayvbdu2aQT9Cg68CGiCsZiQRVATrdrGjRsrWKMWrQiUj4D4cn9ywycaMlF6+Wil31BN2kSRG/XeB891QdJYKCkCikAaAdHySEBWoUx6J8WmMghIuI6MCVm0apmAuJWpTUtVBMpHQFyDxJdbfLqVSkdAhbTSsRr3SckzJn5qEjVZSRFoRAREe3bNNdcYXZffNSjt9M8CEY4zJ0DlgEZGUJ7+lmiNikAaAXEJEtcg8eVWKg8BRaw8vMZ9WtJZSIgOiZ6spAg0GgJy4nDnzp3Zbs9UkvC7774br732WqPBP6a/YvoUYVlJEZgpBD580QUQlyCliSGgQtrEcCv6lkRNFsdIiaKspAjUOwIiBKxYscIICSGhIarBef3f//3fkUql6h36cfsnWrSMNlMOb6hWbVzI9IEpREBcgMQVSGniCKiQNnHsir4p0ZPliPHb3va2os/pTUWgVhHIZAiQUBCtra16MKDKB3L58uXZmHSa3aHKB6vGmyeRD8T1R1yAlCaHgAppk8Ov6NsSRVmC9UnaCyVFoJ4QEG2ZRAsXLY0IafLvTIvm2PgAACAASURBVJk26wnXSvZFtGhy0lYENBk71apVEu3GLVt8suWAgLj+KE0eARXSJo9h0RJOOukkI+2FMEUlRaDWEZDQDmLWlNOakhBdg9LW3oiKMC1jJ6ZpEdhkTFWzVnvjWI0tFhcf8ck+44zTq7F5NdkmFdKmYdgk7YVo1CTKspIiUMsIiFAmJwaFJMyDas9qczQzYycaURnTavAjrE0ktdUZBMS1R1x8xCdbaeoQUCFt6rAsWpKkv5DDBBJtWUkRqCUERCjLOJ/LyU2NwVVLo1e8rSKgSRDcjKB2yy23FH9B7yoCeRAQlx5RRGii9DzgTPKSCmmTBLCc1+fMmWOogiXqspIiUCsIiICW0Z6JmVO1Z7UycqW1M2OyllO6YgZVUgTKQUBcecSlR1x7lKYeARXSph7ToiVKtGVxqjxz2XuLPqc3FYGZRECcyjNBaUVI06C0Mzka01O3aEgzvmm54z89tWsttYjAu975TkODJi49SpVBQPMzVAbXoqVK1GU5nixOls/ujBZ9Vm8qAjOFQCZq/UzVr/XOLAI6/jOLf7XXLj7Wn77yU5oovcIDpUJahQEuVLzEkZE0GSKoPRF5stBjel0RmDYERHtyww03GCmF1O9s2mCvyopyx1/MoWLiVm1qVQ7VjDRKXHY2UNEgLjxKlUVAzZ2VxXfc0iVdhqTNUFIEZgoBCcEgJB/jm2++eaaaofVWKQISqkN+hOQUqIbrqNKBmqZmiauOuOyogDY9gKsmbXpwLlqLpM14xzvegWefVdNnUaD05pQjIM7iki4oc8JPDgYoKQK5CGQENLkm6b9EkJck7kqNh8DZK5yQVE9iCVKaHgRUkzY9OI9bi0Rn/ujFa8d9Th9QBKYCgdtvvz0blFYENBXOpgLV+i9D5kpGaJM5pFq1+h/zTA+XnnGG4aKjAtr0jrkKadOLd9HaFi5cUPS+3lQEpgoB8TnKpHLK+BxNVdlaTv0iIHNF0oCJ2VPmkGhilRoDgfnzrY3R0SrrpQppVTYg2hxFoFIIyKGATLyzY8eO6eGASgHdAOWK5lXmkOQCFYFt/fr1qlVrgHHXLk4/AiqkTT/mWqMiMCMI7N27N/sh1YC0MzIEdVVpZg6JyVPmlpIioAhMPQJ6cGDqMdUSFYGqQUBMUvv27cPmzZs1hELVjEp9NSSTUkp6JWFc7rnnHkj6MCVFQBGYPAIqpE0eQy1BEag6BESzceqppxoHAsSHSEkRmA4EZL7F43GjqqNHj+LNN9/EOxmVXkkRUAQmhoAKaRPDTd9SBKoWga985StGmAQJSpsbPqFqG6wNqxsERKsmP0Kf/OQn8ac//Qlbt26tm/5pRxSB6UZAhbTpRlzrUwQqjMCnP/1pvOc976lwLVq8IlAcgVtuuQVvvPFG8Yf0riKgCBRFQIW0ovDoTUWg9hAQM6f8KCkCM4nAhz70oZmsXutWBOoCAT3dWRfDqJ1QBBQBRUARUAQUgXpDoAE1ab9Bx/p/w11Hc4ZyTjOWti2B27Ma/nXvRqkh+5KPfRfzv7IHmHMWNv/sMnhZZOjWW3FxKAU4vTh2+4qSy6rIxPrVw7B/NgLrtdfDv//f0u2yufDEAxfDJRW+wPt/E8E+WHD5P38Bfasq0ooyCs0zNiPetmLjN25A4Mwyisx99OUn4GoP4Uk047pv/B16xi3nDfT9f358asfgiArnnbQQztVu9Fz/PjiagHzzoLQWvoHEyylYT347RwBIvfyf6PSHEdyTRPKUixD75rmwl1aQPlVlCARv+e9Yv3XkvMlt4vLrOxG9bH6VtXqyzXkN8ZcB+8lvLaOgnHfKXp9lVJN9dHrXdOLJh+G7M4ZIoh+2DZ9F7LqTJ9JofaeBEWhsTRqFs1ZbM+a93o8Xd8Vw19e+B++DxyY1HSzNVpZpxVKbfHZnlmJP7aEAZoV7pZ6umvxIWNDCcW1pBo4fPYLtWzbBc+dvQHF8QpTY+Tja/+pfYf/K80iYJYR7H8Jd0SM4jIXwrLbNrIA/oV7pS42JwGuIPLgZrvX/Cs+WZIkQ5HlnjgW2U9L80z6nxGIm9Vil1/Rv4P9aBI/sTyJls8OzbOa/CZOCS1+eEQQaUJM2jPO8CzYg/gX67vxhLzr+phd37U9h+6anEV33YUw0zbT7c3+N+OdmZCxHVXoMoR1HgJPOgfd0ammqoUnjtuHd6Nl8C3rkueyuGlhz4/+L8Lq3j/t2RR9oW43wt8+DA8cQ+Os7ccOuQezb+jQin3/3hOZKbGsED+zhqCzLtPoNJPvTmpemlW4Er1SfsoqOZ4UL9978JQzdbExkdH26B/+0H2ha3Y7k/3ivoTWtL+pHeEsUT9I6sbTkjuV5Z8EKBL+7ouQSJv1gxdf0INd0upWuKzYgcF45GsZJ904LqBMEGlpIy47h209F5+qFuOs+CjX7DyD6B1orRSYYOIbgt0LwP3UAMao7bG02uDxuBIqYREeYO//ZAt8lfXjw9dlY9+UuBC880TCN2WgiHciYHcepI/w//wXnb+kng78IPc0xmi0X4Yo5z+Lm6CBaLv0sEhQShgWaJbit77+hUzTqr+yh2QyYd4HDMG2GS52wA79D37ceRcDo8yCsp4gZmH2+LG0Gjt37dbz3W8TpFCduW5lEIJRA6iQ7Om+8CM6nQugMxRGnJqj92g3ovTitwUu+8AQ6e6MI7zqCZDNNhW1t6LzeDe/JJ5bWqtxdNQXqwNdCbN8RJIirzbYQrksvYvtOhU1Ke2Uv/Hc+ip6nEun7Ute1l6BzRT4GOfwBbb3Uh9jnTy3hAzofbicN4uwLqIFNDBToQpF2Ju76ujGmBu0K4dQP78F1y/ZQi5a+NLC1DyccvQQv3/5nsOw0sdtD7OZY4VjWhq7rLzKx+x38f3UnvsBxXnOFF85dOxA6xYOgLVTmGHGe3xVC99Y4YkcHYTlpCbxXeNCz7uQS8CjQf708FgHO47SAln/cYp8/Half/RIdX9vBtduPFDVLdq4/r8+D7lXz+e7we60XcA4cfRq9u/phaXOi58sf55xglcXm/yu/hPsTD2E71+d11zsQ44Y00m/h+vEieN2pac3tOOsfj98PyxdjGDiJ6/+KQfRt+QPePWcfNnEOCr14Xw9O2CVz10leF0JHr5j6UrA0N8Pe5jDmbntrv7HRkXk74p2uQXjHuCO8huiDj6KLfCVKNwAueMb+Owf+61fASVeDbHvoGLDx+maE79uD2OvN8Pi86DV5VrqWYv8tbU0XG5vYmDX9n1iJg3jq9XS92/1+OI5ej9iV70Sy2JrOjpEN193YhsSW52G54iq4t9yJvyZ/aOF3gIZyBKLmuN/oQJTfqAA18KmT2uD/8gZ0nE6+WmweFINC71UdAo1t7swZDlGzp4kfXtNfLdr7PazfFMOT/RQ6LrUDe/bgHppEfY+9VtpANi2hilseHUSMH1mhGD/u8l2fx4+taOtKrWNgRxjXhA7gxaNvwftWptt6eD8FEZaRomAZk8JtFKhMl4dklMzxdQvcq+1lfGjfQORbffiU0WcrPBcsodDKPt8pfabkmksizFJQ854CHN4fwxc6v4fek86Cr82CgaMJvhNGSDr6yn/C9/ch3EOhykLm2m7rx/YdO7D+78NkNeXSa+jz9+GGrQn6bJ2DwLVnwXb0AB64sw8dj8uY/A6Bv/8evrD1AKVsCpaXLkRy1/O44YtB9I1qvtScejKCXmo4MGcJOq8oRUCTt36HcJQfCyGay23yoRhDxdtpW72aQpmpT7G14aZr/wyXbbgIlxNLoSZ+dG/bYIf1lZ1o/yKxIwO2LDsLHmL35I5IXuy2bwridj6XMD8KRkEljlHk671Yf18MsTlk8sTUfjRuzPPOnRo+YczQTvGFEeM2sBedf/8Q7qHgZb/Ag4BH5m8M/8T520tfr1zatycBh+csOOcM4vCup9Fxn6RlKnX+H0Ef/eXaOcds/Ulsvy+I7l9J6WWs/6PPo+vO5/Hk/kHYP+rCmpPSrZvHNW7M3V+F4f3K09i+fza813rQRT743FMRfOqLsu6b4d6Q55082KaeDMHztafxCDXXjgvOgqP/AB7ZEuS1X42yDiQQPdqGDmczBvqP4EFu1PpeyVNg3kslrOlxxibfmv6ba51Ybm4wWylcdbNtKHlNJ9BLYf1B8U/NafPhPQdgWe3kuKewLxrBxZ07AP7tbh7EcfLhzm89z+dLnQd5wdCLVYaAatLMAbHMmQ353g5QoEp/6F5Gn5gLSauopeihk2/49bihAQlt3YPkhe8rYSjfDveyhZTEjiC+nzsd7lVjshvEbLhWL6HwVE4dzVh3oxfdK7mTpIarBfRdMoSzP4ONAuBxljrP2UZznNAbCFNQGaDw0e4cpa1KRHDumkiBtidy+rwBfZc1I/T6v+LirSn2OT6yz6c40LluBWxzorh9V5xmVTLIy1bAZYuh+ynutF9PGcwlyeTLIVEaNZ+FwJcvhocCb0x282Qofb+iWZmm2HLITof92yjd2kTgO5079B1P48loikmeWRs1e2F+NARf+0onOihkpl5/nhhZYDEVV9m6Xj+G3vuep/+X7E7d8JnCbcG27N8Bz6cpVh5NYp9plmy9wGloKfOZkou10/a5FWhvexR37eLLBm4O6gFa8bstYTwg7T/lLHSexx03NRHhDHZfXQ/P0ScQo6bhuXzYzbHj1q960M75ltpi9qKkMfoD4sacpJaH9fous1FIm22MmYNrAShR21kQOL1RFIGccQPXtJsaTAfnq8vDtcRNQHhLHA+8nkCEm4mOnDnautKFjosXwrJ1B7Y/xXUmc3LgSPH5n9VIk/9wU9BxIfnSFgqB9JmK7KIv7un9465/a05n7B5qq67gZrM1hcBPItjOza1t2TnG3MW+Jei43sONjA3edafCviyBANfq4UQCsT+ciPYLz4HnvlHvvGyq1rJ1kI+FYsYabVp5EYJfIM99mA74/j04vON5RHA63Nlnl8B3xQr4uHH1bz1CX1zyH1k7C4qgX9aa5oa32NisGr2mV3BNN2MbLQjPcePkWE1+deZbuaafL76mTWHXaHWbC5tvPAcubnyD5poWIbjzYjvsT4XxCAXtJsH74vdy/HbgAX6bBvrZ7/HmwQx7jhQZEb2VBwEV0rKgyAfJJEMb0Y+4qKmEyFhCT5K5mBq2AZ7Ukc9aLsMynxzzj2OlHS00o4rWKzZgNZgtWRk1bGJ+OzBuHdkC+QH1k9kZQliznQz8aTxIB/boy6/BJqY3EfxWiuBHGogjSMGlaZkD7jELUpxlKbTIcxSk9h3NFTFSphZRzIjc9fEDnf43Re1Yus/yWi5ZWG+GpEwRdnMpSaxEoQbu/GJPyXZdhFWhJOK5J2xHvFXoj7eyjzZq+3bAf18YPrbdKNsgjh81l16nBQ8+Rd/Cr/XAtYundWkq6syYVXO1EXsiNBHImNvQ6Tt9/LEUrOizKDi30ATlWrkagc+ltW9jhbRx2plp8jj/JjLYnWRq7PivOFQ/J2bWUdi1XOBG16r0V9zQquZQ8THiRoKHFJqiCRzfsQnOjU50XcuTq2eOmTjjtFZvTwSB3HEDTobHeQBd3+Lc5CGSBAWv9PweHLMRyKyz9NiavGu8+Z/VLHF9N4vwTUd9WeakJE2Sst1Iz6vC63+Y51H7fC0FSUMIGrsC0NoGzx6aKTdF0H1nP46/braR/+Z5Ot2IMf/lhtnkT1Y21Kib/wqfGZA1kKsdF1OyCKG5rhFjyht1oZw13VT62BSrdtw1nRXSZtNkS5P2mTJOwx1Nj/uJaf4tFZkm9BF8d7x5UKyBeq/qEFAhzRySFJlHmiHOTi/2gVSWmTy5qQ8Xb8oduzIYDc2ahkAlO8gENWnycT3FnjZLllMHfToMwUro7TRrtgEPUjiLUkNn3SMMUEyrpt8VTSSiCXGubkv7aWXek39tTgTHhOAwH8hpT+4rxu9lMdfhtzPO8Dgao9lxpPiQel3MaeVoav6A3q98DzdQuGoSLdEVdvps0BxoCL5Cb4fvq5+F5VtBdG46gCdDIZxP5/7LbxT/uFz/qkGENkW526b2kabHjtbM+0X+bbsIu42DA6XQeO0spQx++kyNnTydHvvMcqVDsoHdMFmas7OjtMJznnJc5kOkOcRQAdz1R5/GX//N8+gRP6XPv1dDgJSNZnkvjBi3P+xGR+dDeIBrVzRl/tX0+bqTmuJcE3bR4kud/znzJre8ctY/N2NWU8DL16TUC4/C8xUJ79OMNZe60T5nDwW2uKHxL53IZ8f0nfyZBYjQlxxzr/SSjSfLWdOTHpt020pf08R3Tjm8Mbfv5c+DMpHTx6cRAfVJM8HOmHyopoJddjNNpraJmpN1X/4ihrbfMvxT8sdayjH90l6nSWFH2neshYKb8bGfcB10dF0m+0qa+Z56HlSCjPBHi+ygOYDOwe6VZcZhkvaYO9Esc8zsgMmUJyIGWJpNweKU1diZi+H2LyF4XplMSDSE9E0RkcV7/Qb4LzsL7lzzgIHpO9H+uf+GeN9VuHW1FU2vH8ED/k2mz408IERzKH0QxbydooZQ4JtSKqWdJVRoPWl4D2VoH7KC8lgGPpGxGW7CW+G8eD2ifddjMwXfFs6r5xhipP0HkwtHU0IXG/6REeO2J5Y2haEN3XQN6PS0lS8klzT/hzeZI4SgKVz/MfIg2QSBJyh7P38eXQ9omi17tMlzTEFQNtEGZdbAOEJi2VWN98JUjA3rKGdNj9ekovdLmgdFS9CbVYKACmkyEDyJ17Mj7ZcDMWUZKnyaAgwBYJAmSbn3BiRA5QlrboZl484RzpzFx/LtcIlfGs18IfqJHTbMkjaTYU28DgdNm/NYamwrhTT+O+yPRifYp9heOqR7S9EQjWg8BVSDKdLMYPY5TiFGqOkk61it3Ih38/8h5lLDt56+XHFRVe57HE5ieMKa29D5Qv53Cl7lzjltKhFtpwh4GfOM+cbAbrSvvxXWj7HsxOno+vIl8Br9ocZxz8jDHo4NHvhsNJvsehS+H/yuYJUTujFeO0ss1Erz5gjsiGHaX5Lmz9HCaYlljn2MWr+/+xdidiuc9/F03XXt8K8U4ZCHXQwzutK0IZCZN7Ihks1SPzcQ5WiLSp7/sr7FhJbMunBYTxIRavLrv9DmTjZDJocdA+dYbVnmkbdynqdFu+Omi0mCa8DQxjEeZf5DO2OKn5oLkx0bsxXTsqZLngdTA42WUlkEGtrcKT449stnIykL32CGdM73ucy4Vza0UxNz+6YknruPWoX9VkSNyPMWeC5tM/wjCjGd0UPmpEAlfmkvGic8GdTQmQkHMX4do8vK/G2R06Fznqezrvh8zcbajD8anW+DNP+1XJoOvVHo/fzXT4aXoUj+adMRPMfTgr5EM7V/Zp8vsE9gJ8ydo5ParuY9eKSfJ4++uBnB/j14TioXJ3WabMuitzN8BwWrR/YzxtKmJxh+II5eUw1m7LSbrBQyUzzllGIYkR/B2kZHapExeYBiGHOzxuZT0X1tG/oYDuXJbz2EntXXjHDMLqtdox8er518Pu3rx8bteRpd9y6E/8olo0shdnJqK41d11d+hCBNxmnsHDw0MebxCV5gmAdiepx+fM9tfZTH+znmprZSfNWUphEBbhAdPIjzJA8o9fTupCDFmI0irImAMMq8nbdVZcz/7fc9hE6eIu0zfBut3Eimte4TW/8Wmj7Tgv2+HZxDyxgyZJmNm0jy1T1R+B8mr9xyINvktFA29p1ANmbgcO/cDCHUEoricDSM9v95AKmn4sbNlpVnpbOmDD9a2d/GHZuM/27umh7bpHHXdNZvcOy7JV8pYx6UXKY+OGMINLYmjTvVfdQYHaf40cp4Whu/7EPfhRmH6RPhuvYq3E+BbNVJ/QhujSNps+PqG33ozT5T4rixbFfGofWUNrizJ44mUQf90jxmyIbhgwg848CTlVH2x83ToxMh17Xt6T43J9HH2EQp+s9dfn07+5wvzlgJNTBAZe9XGV7CSW0iw2H08Uj+cvqBPfzPPNlpqInKoZPRdeNqrOHBhwR9p3qPOuG/VMRlOd0mPoQn00zkxXWsy7I/in9iDDcJXXHrP7enY8eNqsp2IUMDtPHjwo9ip/+XiJfTlKLPjtdOnva61IObVi40TukGecIub90L3pfFLsVTsn0JC1bRV+n+r7onFEA3f5NPhJtjfptnCVr7KRzwxGu8eQnn+VXlz/P8FejVUhE4mfHOrnVgKcMpROlLGeWJxi5jIyNBUU1zX9GySp3/FqzlZiwmsbV4gnTVhkvQZaZIm9j6pw8U45+ta6Pm9+ge9PJUvIUhJwIX2NAyh/Obvmiua93mSUzz1KX4j456J9+BAssqD0I3OrGWOEQorMWoz1/ruYTXSg2ZUxSw0m+WMDbVs6ZLnQeld1+fnDkEThgiVbL6jRs34o477sDBgwexaNGiSlbVYGW/gdAt6fAYYGDcnTwM4ATjc/3dv+JT0TZs/uGn4NUDeg02J7S7GQTOO+88bNmyBfPnl+mXWc8QZgOlWnA1c/X2zniu3noGW/vWCAgcOnQIixcvBhatpP+lZ2SXd/2Anja7MVkRq6HNnbU9iZI82Zneew77ozFH3EkOXO6jmUwFtNoeXm29IqAIKAKKQMMjoEJarU6BV+II7ZHG58RHo7nP94XL4KvVPmm7FQFFQBFQBBQBRSCLgAppNToZUrv2MOK2UCYwbo12RJutCCgC04NAM3PsMguAlxs7Z9afdXqq1loUAUVgYgiokDYx3Gb8Lct5n0Jq+4w3QxugCCgCtYIAY2d5L2PKJiVFQBGoGQQa+3RnzQyTNlQRUAQUAUVAEVAEGg2BhhfSUsl8B78bbRpofxUBRaCSCKSSPNRTyQq0bEVgShFgqBT9Nk4pohMtrLGFtGgX2v0xA7uwzwFfqFJslDHHvExcLakB6phSwXY4OsKT6mGkg4Fag/nGIQa/k4m/zeC1IyqJ+Rnk1oeKDd+kejTy5SQxsp3A+G7xKSy0kkWlgvBaGSSYgXWd5o/bL96QHA9XgfGoZHtqsexUGJ0dfeOmHys892uo08leBo7uMrKgFKV4Hzo8LrjcftO3tujTVXQzgg478wbn62CK+W8dPoSrqLUTb0oIHe19JQdsn3g9+uZ4CDSwkEbBiQKap8NJjGKIxGxwOcvPLjcewOn7UYRjDrgdpT09U0/F+zrQNQlJJxaOwuEWPCdKcUSiBcYhGSUzJ1PPFwTf3oFQNACPOXyT7cdEWz/+ezH0BJJwrmKKrHzC5vgFTP8TjPQec/JDymC6UfMn3OVi9GCOR5KCW77xmP5WVnWN8d4AUu3t4+ThLDL3q7p3oxrH+RLnZqo4qyPv7fQD3WFEwl1c1QmEOjsZrLkWOupks8PcMOZpa5xrxD6Vgabz1DGdl5KJcTcW09mcRq2rcYW0aA96GKyi3c6hlw9OyplfAJiKmSGL18bFWykZcCrayOUY7o3Cap9oIxP8kFso6KYzAEyoSSn58DOdlYzJaBJhweHKz/wtzONny9Q72X6Mrnjq/k70dSHk8aPTQVNCIp+2cOrqmqqSEhTM4HKNzdsak/Goow/SVAE2uhxq0fx9dvg846yLYnN/dJlV/HcswpwALmfxFHKpCEIxJ7wuk9fw715uDiU9WfUT8y2zofm4ZDISBtXNRsrAuqBUQjVpVTCQDSqkpRAK9NE050kvKBEAmGoo0uWDl+ouh7uLQps5OskIAu1ufqeY05NqbndX2Jy4CfRQa+QL+KkW9sDpcKK9L555CeEuDwUWan5cTFgdCCHJxWuLB3gt1xTANDweNwKx3JmQQrSnA16PB25nbn283tsBj5ttYVkuX182lVCUmg1vV8Boh9TZEaTmzs++0JzgcFHLNN4ONUHTA8vtDCfQ1+GGT/pRsN+8FfYTJ+mbA05PD/WQJPnIJNgu2UKnqDFi3T6j4jBNAO0ISNtdTLuVBOLBTv6eft/Oe70Z2DgOURmHjnbet8PGexlI4xQWrNYYutq91Eja4PCa9XI0xJTcSVkC+fqRCy29gqYd20z9/BD5e6w0e3Ee0HyYTmBv3gx3wNXuh98n/WYg4q4QokGa4r0UgiiYdmUnIwVQv9eYiy6HHQ5PAFFjniYQ5Fx2cwzdnL82izud0zQVR7CTeBlzhnOkJ5r2i6IJs53Y+gOd8Hk5Rxxe9MTyCY3EKxyH0xjUkRTnx9jK+W98rDjefWY9brbN0xkyd+BiGqJp15zf8R7Oz65IQ/lmJfs4Rpz3bgMoWQs5ZnnhB+6eNFZF5n5hfMeurewoJUJcKx7yC46vM2eNGddlnpCfOT3wh7kgjTHso9DUgYDBA6kh5dwKR3vRKeuNfMgTkAU2HjHANueqk+ruVJ8Xdl94+IVkH9pdndj+U7/JB7ge3F78S6CLf3cgSH7T5W7nPBxVhzm3POSTshY6ghlmVogfCl8mL+PcNtaP9D0SQU+Hl/WQX9GEFx+vG7yfl8fxejzAtRcwSyDPC3D9OGU9ervgD3KtZMwlceLM+o21R7cAg6cKFexPobXNPMTkb76eAHwcs85cS0ehsgqte860krEhO0jlYwnpXuh/pxMBSQtVSfrbv/1bSTs1xLRQlaymvLKPbR66fOl1Q9teS7+297ZVQ/NW3Tq00/j75aFvrmkduu4J+f3Y0P3rlg9dvvnl9IMvf3NoTcvVQw/LX689PHT1vJbsvde2XT3UuuZuvj00tPeba4fW3PTEkFGcPNfSNLRuM/967f6hda2sN13a0GvbrhtadfXDrGWYjj189dDStd8c2mu8y3a2XG7Ud2zz5UPLL78/fd1oY8vQ1UZBx4buXjtvaOnGbUY5r/H9lnnLh256wvhr6OGrW9J1y6PFaPetQ8vX3m22pVi/tw1dt/zyoc1Go/cO3bZq1dBte/nrzpvM918e2nz1qqF1d8tFkpTL9lx3/+40HvIcsd6dBmdo23WtQ6uMAoj8N1cNNbUSX6Psf6lNMAAAIABJREFU13LuvcZxaBpqWZfp/86hm5YvHbpppzzH9ixle8yxNOrL9sMoNkszhi1b8PLdnBO37TbaIvNtebrxxt+7b10+PP+O3T+0tmken91p4HWM77Vety393jfXDC3f+IQ5RrtzMDBum8/I3JN3Bb9VQ2u/aeLOcte1rBu6X7B9YuNQa8vaoW+mB4FDt3xo+a3ptg2XJL8JzmuGvrl35FUZm82XL82uEann8vvNNfLatqGrW/mO8Wd6/q25++Wh1zgua9feNpSvltGlT8Xfq1evHjp69OhUFDWJMmSdci1mOk0MVnFtm0hxeV/OdWswmiJz/5gxjnnxHb22si3dzXXJNbctzVlkfqXX2M6hW7leb3oivViE/7Suui3NU8w5cfdeuSfrWtbb3el7e28bWrX8phLGLmctkh+2rNucXvMscfeta8gT0j2XOZ07/1/mvF6Vd/69xqnKtt+aXgsEYqjVXNsF+SH57eXzWrN8eedNS4ealpKnGFU/MbRxaWZuZsEa+wvncF4eN2reb758TRbj13ZuHFraZJb9GtfNmuH1ZfDG5eR5fL9Qf/YWXNtPDF3XOm9o1U0PDxlDk6VCZRVZ9+VgI986o81KxRAQuUbkG6aFGsKff3nkz8L3GvcmSw0ZJy0V6UPEyUTpaTUAtQUJeLo7c8yRNlP1boWnJwxv1pRGyKlNs4sUHafWx9GFoNdmyNQW+Z/NRs1cFN291JiEXGktg4UaD2oTLGLrtMj9JIxDMxZqrahiaA+Y2jyjlCRCrM/V1ZOuw8KdUy/f596vl7s3X2/GryXdvrCx0xF/Nze6+9yGVlDSjFs83eh2pf9CiqZAq9HRopRW1ftNVX2RfidoumWZPqNIO9p7e8EuIklTKbffiHZRa+boQcjHi6RkNIJke4DaSEcaD2IWDluRbpIUYoHdsHOktTZufxBpy5CFmjMrLBZ5hlo6mke6Qjn9ZwFxqcAwJVPbZHZxZD/kgQzNHLagyau7x8b2O4zG2DhPElFTg8Exj0aS8PrN+SfbV1s7uqlxky7JEFtNO5CtPYgwMZGRFaI7P8sy/5B/Ytxtcx4EQnyXmotAxMV6TdytdtitMjuod6NW0kLNXYcjDRqtxUabxhDdAKLxGOc4E38bN61o5/zscnHORU0fy2QIgaiXWkLzfTE9c2yMYWMP3NQmdPT40UWHI18vtddjKqnjC8QmmODaNDstayHl7M6ajqPhmOnDWWTuF8F3zNrKQEkNWI+FGnR3eqbYvQH4yXtSoW70Orqp2TPH3cb5Y3AMWUbUVPs4f+zmQuJIdVIbb5cyZU5y/mTmXcERy12LVhtsKfI6PmxLBtEd5PoNp+dINBKnJjwzE9j3SAJOX56ZwXXTE2QA3qipsXV2oKdbeGgcgUL8UMzwtFZk+LI03UX/N092erNd43WkAI8zeG1m3sd7EUhS029ibCGPtxMs8ZlNhQIIkc9FzfVF+yj5Hflywf5wpRRa2/EIolb22+9Jj0UG/EJlEetC6x7lYMN1H7dR619wsPXGdCHQkEJaNBihqagnLTQYQg79I9wmczLMdjzpyQUnZKGZqs8for9VHAkyobij1/jQJCMUSngyyXyMTI6+GGL+ifXyQ+pBV5YRxBGjv1uH8aB8wNJCWirqR4+NPkq5vEl8NaI0S2V8NbhEXMJdEr0IJb0IZCojk4rF7TS/sshYBDE7zRZmfXFh/DQNpHsTSx9YcErdxUkYpzPLOAv3G3YvOimMeWnC8ffRq8+R7kCY9SLRjS47+0Sv2gyrj/G6i6bPzN/yS5wmoL4wsUwk2HwKfIaElR4HX5abphCPp9gX9j8eJKNqR6f016A4ohwjD6tOhWh+JkPM3Brdj8wbmEFs4z3diLXz1Js5Rhb5SrDv8gGz0lhszL/MmPNDHnfRLGgClu5PuncWmitCFHjCHKt4IkZh2YNglotG4e+ggEshSvYDqVAQMZqRXBkAkpwznLntFNRkU+LuyNwRIZHydXeeL5e4AbgDiPFEanb8pDz5iFlpeuIrqSAFRzL0Dk84/Qz9WKIpN0KZvrIv9vYuWqK5IWkwjp8Khygw+JBZfrIWHDJpDSJO9OF0dxaf+6lwYXzHrC2z5Dgd2y2e3uyasDhoduO9CE/lOukmkB1LjluC69lOMS3EOeEyDlGRyAPDSTe6zKamRgmXZjVj/kn93/beH8htZMkbzLm4WMETJmIjxNAYDWMjGntOw9qG8eZrnCVaJxh7J36WuJZoveZaorVBa0VPuPmMwWdciM6FGPcMYccRPWEtwROe84S1hGfsCBvx4glrCW8d3C8LAP81yWZL6j8Ss2YkkUBVVtYPqGRWZlYl3pfZXGTFBIHn/I7nvHECu1urdxqbtLA5yBk2XOD9jw2448+Qg0iA3F2QbaRbcFmi3hZ5mLFchmwyFDnGGAvOXv3i1e+tgnxb2SDjFt/7PJhSgbkxe6UZy3rDRByEWMOn1K53MhV5gvhNn1qbxgNeNs3tAr9VuYMF3Cq/G2jxfFw/73na7o6NWsghJKKexqu9y/crRGAPY9IwcRMNcTj166cmLuIGmokLRSwxEMuAh5BNOpjvAZQx7Hr0xjTEqsl0KsmlVoP1Z7Z3sDXEwlY3tbqF4jJ7uZOKXtUKFjBW0vADHYxi6iBubWkSYKWjdgetCBEWfini0GqZWQnROv5L9Yd4o0pY5BCA+MFtgvfZEgJr3Pmx/JXgbHa3bhu3sp5NoEz2CsSUNPF01e5YE9bBwlhccVc71pwGXI7LQpshFKzOAEqqD2EKwas2CvBzQCzVnFemCWsNHgSPP8MYuZoqWF0m2IjBZCtrRIPM8jia6urf68IWK9sBfhwT350dYWH1EKNY/4CpseD9a545WzRaiD2pXgEeD/+A4VviI76oT1HLpd7Qo3EfdYBJNXK8fwNYTvAD3Pywntlpy+8QYnospRQu/ijWivzs5ZqjluBlmvMyv66eB2hxEw4Ud7yQwumUpvwHyncGix7PH54XkedTipd8H+Nb0oSD6GtLJhYWSjmZzQUoQpjDaqG15d3fjO8KvdnjAeZQuM7GEXJ9vXqX6rr8jhjqXeO5Zqi5pgor5+pdqb/iGc9lXX1xzT9LcxGWNKWkwRI4Yitao6FDJi3tCmYZBRVSGf5XilIUEO+1qttvloeVRdJyjIoSK5tNnCyuFAjsX5Ijqx3Ovq+TcbUcqt97jtU0EavXFFYOdWiabN9nLIdYPKr5gD+8qMrGbfw2rB/PtrmtFPs1GuwmbDbP+4tgw+9QCsdI8wZsBEpuXAECe6mkJVmrEo48cZUCAFN1DXblkuAdOggKHWbUY2sRBK3RwourFCAWGZUy4My0qfkPXwqLm6Ekb1VSWDR4mVjJoMqSliDw3tcHsELNqqkPBYR6BiXHqL5RgrbsFcvTFKua6qqqh9V1YiOAHJ95Es+CVdkaldZuKK5Y/zCv+f1VPcwKC84Cwa9KGm4ZNwJVp0GM1bFOFlbGvgvLDpTTancsXLMI1tVhNZrikipslczxQ1RL2TwYwd0CK1ofGy2MFmlwGfMOQeaPn0Oss/ulLslUKSTQizFGCEBzLqozWCgKWAvhNFQr5dmO0qVxNIRqVq4J2xiWr3wEIQ1lvTnCIsbzt2HZytS4o4X3j4VjpewrrtV4oMTi3ZuORqQNJjRCoLIJ7DK0U5ZbphENqBch6LrfPOnKCtkyqreOSbH12IDrUVM/ingmzS12p6zdeZzj9WFeDMXKYuF3zlC/6Dkl2LlgLPST8jOudjMg+LpP/QRB4Z5NURAqt9o+lTTJalc+Rr0yFzC54Q6rFoOb3/0t+K7SmwELC2s6l2+UwROAH1y2PicZZMtsGqVQqDUo/riAOhEUpeadYAVAny0Udj0aZGUustub5yesaHl3MF8Es8utHrdimWXUgvxdfD/UTtHZ+5dSiHeY58xmeZhijsH70Mhl7suYK5tM71xlc5OMQ7/z9x4egBQLfaPhFnIQmwZMtdoE/klr+Z7nwTKJO2vHs21uNzvmeZYvl0202Puwdt7j+e+MTRFSAI9OZ2a5WO1dvl8lAnuopPHJ31jl1T9SauItrFR4NaiUHuyQSTLERan5gR+9SZ9GESwQ/DvIP54LygcLObYCVe80H69QaymIW+jjB3q+4gI9/OAGw5jatRWt4NXm4q8XTA7qKw577GOXX17PzxwCX12HcPYQy+YO2KW5IkDZ9bnwg8sTufoxPeeVUu1qRXXbuOF2HIzg0lDkWBgZEPJKi6oEr9FBvFKEmLyk6nBlRZ5B2WTg1ZBy1OtPsFKuFFi2IOVZpbjgVwMYBWT2O0rQV67mevw5jjTwNeqzhruyUlau32Yc6OLascVu02GAuLBVPx8w0GFN5VM4lle+eK9q62EFMePKFiucqwbhq1cvI4btQSHOYA02MEjEiPVixHv1a6taBT2/s1k6fw8HIZRoFz/I6kexUoy5ZuXKWhd7wu7nypLZUKz+hfIWz6216IUSvJuqZHg/hhFiFPGE8bk3yGkAa6nltMlkt93ie75M9Bv8hnmM2E291Uxgft41AFAEfCjc57/76imux3dpbrHMgRxpUGQZUgkLuDh7NE5qHjC309mjGtC4hd29ahqtLhRgiWuUI1YG2RplnPOIVuciy1gNsZgTc25FA4mlXcH8HWEiOixSNYdn5myTESbHbuc+4hqbemvlIeRyiKVrtdhkMc0vql0vkHc8ImiTjAO6i++9ejK14M6mA5ydBksksORSsAxNq8+8S3QYIO62/r05O55tc7vGvqG78ltxlpbqff28vwA2RThRO5Kb8IxqJPL3dSGwhzFpmOaIaq4Cm3n1B29mt55BzWpjyHWwjbs7wrZnhyb4MTSx4uQAdxULygIScVlW/dQqIVfHQWDVaHS75EChM7BiNNDWqCcZV1erHLiteuoaH8Fh4egPbCJwwVMblofxANu2PdwzqDf2q637HOQdIL4IPoEWVqgWW1R4tahW0wjwNSpGKtfnAOKRC4QKb4cf8tgKChFvMjBhnWoqV02qv3Uc2xD3cRwANijgOIiN40Z8SK+Fs+WcqRJ8Rh8BygZGEcAtZg8gQBH/gY0DA3cEd+iYLICrYawMGRfT7eEPXMgxB9lCmDKeajmK7eehRt0OYpuAN0doGF2cY8cxHXCRhojHQyg9FEL88iAA2oXrWek9/BwWVsowty2Mo0PRtWILqxisXwVcm2cWpGxlgGs4g9DN2BLIwV1cVJAw3rsasFS5G4dAA2PGERs+4sj4GIEWu35bOE4A71AC66WfwViL+wEjh3PYgj5id4YIEu9h4eBDKYZLuYc4QBa6q+dY8aJkrSKfwuLGvNSsVQzy32ytxXEz6v3VsVgYqGNbrDF+lJlPH89Gx3sEPnTEcrJeiBcegdvgKyJqG3inHRy3gNjNOmRnTvqb+sQPsVnk4WOrDYs8AvMRo8fH9ViwGlvKXbbl3Uf79fii1eLcgku9g5inDhZLXSxSugM8765NUwg53YZlU8VkYeOKhzADWGIDKPsaFlRjbETiO5Xbc1TPU3gJFt3hagFWW6PSLc9uxWrF7wZvVDEXrWiQQ2qTQM+YPWmtZVDWx3sK2iH4W5SHdh+bLHB0hj3haQ9LMUIG1Ou4SR4iFi81sSCoqSuXXafui2VlilhjdTOmAWIFcTZHJUdn3ODDBhm3/N7jPQfPHfxpQ7jzbwOHz1QRNIjlG7VwLp5FE8SeajoWaTgknLlobRjPprldhWg0lsDl34pN2Gya9xCwO2KTU+DjkPdheya3F+GRz9eAwOduDz2v/c07ggPbo4/nx2Ccx/+Xvf++fHb/pFy72/zLdrRMjY9fODot6533l9nT/tEWbC/0zN88xrEg9VEMF2p4wco34QiOlziKZuGklQuO4OZVv9Cz+4Cjh45vpszhY4qOHrxYOvro5qF9TRzhuJV7OOpp8Vioa+Lkq+hWjuC4FKWV48Jql+Kl0N9MtIgQi6b1l3d0bq7+xe7wTlRzNDxr0fliPewvIcH2Is8eLvICu4K7la33Ii2/xrqcBePb2TBxsWeXjEeUzXZ03qSnB2teosPaVVvlbhJr184LrP/eFJ6aylNy7ewIAwqB71hdvUwsTk9P6aeffiJonHT37t3L7OqG087gAupQ5iGGqLHH33COhT1B4GtF4Mcff6RffvmFvv/++691CF8v3+x+xZEs/dCTheHX+xSF8x0Q+PXXX+mHH34gHGZL9Dft5RZ/+B3Rn97yabY7UNpcZQ9j0jaDcbl3EKcwDS+3C6EuCAgCgsB1I6C7NInc6+ZC+hcEvgkE9nB35zfx3GQQgoAgIAgIAoKAIPCNIyBK2jf+gGV4goAgIAgIAoKAIPB1IiBK2tf53K6Oa+TfU0cuSREEBAFBQBAQBASBK0Vgj5U0HCrbwRlVSNQrZTMCiYdzz3C+lRRBQBC4AgSSCfKg4hw1nMJv4jy1dj/AaYpVCXG+WGd6BTwsdJH5OMPskmXkFHLYuuQ+rhY16U0Q+HII7LGShpQoXoxDV41PRBOpQPp9wiHYV1NipKfyrkNbypELXGxpV/OQpZf9RiDB8SQ4VHYYqhRiSYJdkgkOvh5/WSGTTro0mF7fnF7t3/Ehh7ufKof3+42R0X/7COyxksanWCMZuPaJDxkn4Y8h6HBo/pWUZDqmDKeUX0cpvp3Dnq4DPulTENgNAaSL4lPx28gsUhWDXGQOyJAP+EuWLEL6tOvT0ZBFb7l/DXINR8pJEQQEgTUI7K+ShjRIljup8t0hFZLZHiCfXpdcJGAznQESW9doZVPqt5HCCDkmTatDkxSZipCrrW0jcTTyTw4cpCtJUBefvQ5cAzbSeCBViDNoclzyLaSDabdVSh+r7eFAz4p2GvRxjWkjgbuJNC2grUo8QD2kc0F+T9sdIvcj0gEhX2fid+D+QD7Pulrzzyb6BN4H4MlBsmQLaaxGYd0S47WRWmXUBT2My8HhhTFy43WQgNtCCpjBbPAYK/LTiY62Arh8FQR2RQCK1xiKloHVoI6URYNpurklUp4ZGtJuNQICNfUO8igiQX1Tihj5gDsuOXCHthdkDHKz1XIK8oRlUlzP9WmXzI6vUoRZ3f+H/l+kMupBiIWQLU5njAy8qyWlCfg1kXrMaXfJi+bSJsYBj+6kEYw5zn1EqjlFICXPQTqkyYhc26FRzBnrkKaK5Z2Smxi3kj1IObSm/6i/QFfhVclRG7S6k6SOieVUTujDH1CX5RQ+94Iva2FcRUK+CwI3AoHLzr1w89JCVSN+//NJeVTnZ3r39Li8ffykfPOR770vfz45KB+95s8fyuf3j8pHr9SN8vXpIVLaVAkzuP3xLL9TVe/Bi/cN8fLkzsPypSL3rLx39KCsbjHtw/KUab95XB6hz7eK9Mfy1aOD8vjpO9X+w7OT8tbBg/LZmyY5x6vy0eGD8mXFhqozK5vol2/KJ8dIS1Pngvr46lF5cPy0fIeGb58czceLtEb3bt0uT56+ARfc973y4NGrGXnm6/6LdR0v8CAfBYEbhsBNSAvF8/r16QGfZDn/c+uofPp2M1gfXp2WRweQOc/enEnN8+rRnfL2/eeQUCgfX0MmHJeVyEC6uXuH5cOXtbx4+6Q8PnoMCYCCFEh3biMV3et5oh+WY5vm9MeXD8rDhy+rvtHH48NbM7n05jHaPW9kwQf0eVQ+VmN5Vz49vlUeIJ3QO3X7bfn4CPcqgQpWT8vDI8g6voWy2r/6XtNlWXwEAanIoP/To5N6jEjnd3gbvNRUeIyH9RhruvKPIHDVCFxFWqg9taTViX4tA/ITn8MMCWVhWZu5PrGibbEOnSLZMHLutqob1iAg32W7fN0eK8WqILG4H9K4yiZdXYI1jamnE58yWOmqW6g39Krk6uYAOW8HVcJ2lcoWScVr32mMxMDtEWLQrNoHwMmuW1g9rnHNbqJfTD0am7DC1dm92bXbAt+FSryekztqxouVMZIXD5GAvSGvz3y4OaUp+KrHXw9W/hEEBIGdEEhoEvxxueZffk+TLdY03fEoCoekjV1Y0wc0XTIWabCst1VSdNJgSTMxP9V9WN8SmzpOLS9Ml1yK4DqtutacPvVnLtTtjCdhAg+DUyUy12zqVhnJtzdSd5G8fdCBJZA/G7AYhjSsBZYGq5eZp2c8AGeJ5hh7DgufXcki9N+xYfVrLPsaJ4ivZa5K+p4iPb0UQeDbRmBPlTRM/NiAy4AlCgs4C27OWkUpYooyk2yDH7xFvX6LfLgL+0GKGDYoXkoOJqo9LPmzoiFGbQLFpwdXhOuOKIUQMSGWQkhZG26BqkDhcdpkMg10lwYeDXpdCJ426GOnqeIHcSkJzP+NwMWVIg6pgHtDCeelspl+HERK2NajQmcJZUpxBO883lp5g8ZGqQ23al0xhmS3lPLKhXkx4AJd7lW+CQKCwC4ItMg0bq1UvD1bjG2ioBkueWFMgZtSHwqOCqeoi67XihjPbPyvnI84JifX9IX4WuQnbkGBy6tGGtrM5MCmTuvrOTYJaQuBunqrtVtb9D+PK4NrV4fC5Y+oj5AKFyEb0Tn9VrexSSnXl+J8dQjcPK8HAtVxNvyaKzV+KYLAN4zAfippiB+LoYBZrPVksFLpiIFopBisVokBq1X90M0ex2t1KOkjLqwJGsvRhtCmlpfZpEN2NyByejTwxjSEgmUito0VwAgKEXS85QKFboQYjiGUwc4ASdf9LpQ/q1IMQTsqat7qVjFWt6bTcLRIagN9tgDGjdJX1U8g9A0woqUI2sV4G6tcil1kLb6uqkEpU8pnDUYSUtRC7Erzu7AyDPkqCAgC2xBAKjhvRCe353UO7vs0cjeoTEUGy/VcIbGQoDzo5zRCvt9Z2dB0GxcXu7cSg/pJWlAM+YYYWMixLjwH48mQnEvn+2KjlNqCwNeCwH4qaWw9MmGCx1MqopAyWKmM+onluMdWK6zfoJxN1S4o3e5TgODdeFovaWHZSmdtYvKGGYJx4Z7kDQCtAooObHCsAWYI9teM2gWAsNk4oCkI5gE2BZgeTfptbDLASjUNKan5ge91/lnxlGE7fmuuOC2+WRvos5s2ydAvK6GqpDQNceSIA4cnxjcfL3iFe0HxyqVWEO36azqF4gkrn1Hdlb8FAUHggghoVp/C9C29evGcXrx+RwkWfBvnUzrGdBthaTcvhmmozTtbC1vR2Jo2q8bH5ixatra2Xrqpw4qWZ42iCJEAf+q8d9a0zuGFqSVTCgibDmr5pmu7bj4Cz1rjwq3YyqG0zq2Hu49DagoC3woCe6mkJdCiWrBMschJEHRmLPgt2WplKStYQn6/2YlZxa3Zrq2eexonpMMHqBaHBStEMO9XXxCD1qdRBGuU0gBZONVCDbuW/B4LYI2yNIXmV7sgYNUb9iekgQc2WFWWrXl8GLQqCHkDbhPujEsBAVxnAdhAv6qW1vEqUPOwc3Pc6hOHl7BFbW6VY0scdqI2RjpYzhLlpuWSUjAh6naqb1Xf8rcgIAhcGAEdc8yF2w8xFEpMbCqw4Ds0pfFsRyXOYsTuRmsWLrGpIaz1JsItsOBSBeerBYTdmcb6+hqY2KT4sWxIsDhVlCB7xs3CFF9bZkvJDyXRsJt0srALfKknvUU6FpCVUbCgZOxTiEaNere5fx1j1SkM6j44hCRqwROxFbX1g5SrgsA3gsD/+o2M4wLDqALnrUELbdhKBS9ll9UjLmlltRqyUDCp2ytw+nebxlgJavaQfJyMzYWD8LM+3J0phI/nIO5iRF24LyfQpExYq3gDgMkkDJzW73YRtIvYMEgmE8qYzzqPiyM/ILTtGBY8A1vmub7SwlgZLKjiTXXFncEVOoWC5dIwwPo07eG+iZg4bDrYRB+ifuBxkK1NAbRHDfXG4w5i2jIK2A3aq8cL1ye7evu1DEyV8jqsfkiwhX5iIHCZ2ZIiCAgCl48AAuVHWOT1IFuwJ0DNw5aDcIguy6ptBW5VxH9FPYQmjDRIkRZ1fH8WsrHa0my7lHcRUoHF2ySALFqooCGedjTtISZ2TDpicF0c4TOtLWstxJZ1Jzi+R8k4eAHwbyM5l/rARqQRMiWwTPQQU2agv7YGD0VdabX/xbZmD0eM9Ho4NgjjRwOjP6ZPPm98deDyXRD4ChH4jresXibfp6en9NNPPxG2qtLdu3cvsyuh/cUQKHACeJvCfkgj64sRFUKCwJUh8OOPP9Ivv/xC33///ZX1KR0JAoLAfiHw66+/0g8//EB09++I/qa9PPg//I7oT2/pc1WsvXR37tdr9AmjzSY0LnrYWfYJbaWJICAICAKCgCAgCHwRBPbQ3flFcPu2ibS6NJl820OU0QkCgoAgIAgIAjcdAbGk3fQnJPwJAoKAICAICAKCwF4iIEraXj52GbQgIAgIAoKAICAI3HQEREm76U9I+BMEBAFBQBAQBASBvURAlLS9fOwyaEFAEBAEBAFBQBC46QiIknbTn5DwJwgIAoKAICAICAJ7iYAoaTs+9mw6wInhOAAS6aBMJD3ujpP6cMaYBhYOvMXhk1dXCpq4JvXjy+wxor5hUD+6zD6EtiAgCMwRCKiDFE+GWcsZ08QJ/CNqEglsQyrq4+DZSXNc7Laay/emOKDb8tLdG2ypWQQuDuxeLzDycZu09rjKZLCFRoiTazvTLRXkliCwZwiIkrbLA8/G1O2n1EW6khgpCpJpn4phl/x0l8a71slp2u/TJNu1/peut9q/RcMIKavkrLQvDbTQEwQ2I6C3yU9qOZNENNJ9GozTzfUveidBjuHRXJFy/BgHVxsXpXLh+npnjPRQnfUZCi5MTRoIAvuDgChpuzxrzmmJNCjIglKVlot0UUgvxdnXv1jJkUc0Pnel+cW6O0NotX/kI20h7QvnppEiCAgC14AAUri1TcqS9Mv1jVRwUTqXW5puUGttbqcv16WixOntWiJMvjCqQm4PENhfJQ0JgvtIpK6ze8Hp0WSbwoW0XsyfAAAgAElEQVQcdjoUtVneY7wYjp/QuN0IHSQRngyo6zpwhzrUC+bmsCxkNynyc/If5MVrXBcx8m+2vQkNkDzZ/R8TGrU75GEFPQIN11vjx8ymil8LeUDbHW/BBZKS5yBHXtq8rVPqmD0K+WsxIRefxz7yhFo9pG7mZMddfLbAD9y2do8UqwXSP53pH3TR14wu49Wux+F0yI9r/27mI0F7j/xRl1zct+wuwRMsRRAQBNQchJzo2si/i0WP2abBNL0QLnmeq7y/qiDhud9FzmDMdwd/3FG4flGX83xmlynmI1ynLE9YLUsx953BFLnR++S0q7k/c5MWkBvI8RssrDujnkXdKS7s2i+zmEfk9TrU5tAQ16O4oTftktEN67GnFPSQnxSyso08xv4AYxrNhUYRI39nx4Vcsag9WDPGAm7hJV5D6kEBdPwFueuDphJeyFeMXKiMhW2ZZHdh0Wt4gkxl+WvZbXK7PmSvSb3ayJjBW9KGzGacGSvkuZciCFwPApy78zLLb3/7W84NWiJ352V2c0Hab8snR7cUX7M/h6fl64+byHwsXz85KQ8OH5Q/v/6wUulN+fjwdnn89G11/e2T8vjwcfmGv318WT48vFc+e1/devfsXnn48GXJ3bx5fFjePjotX9b3yvJd+fT4pPx59n25m9enR+W9htDbp+Xx7YPyVHWCdicn5dN3Tf2X5YPDR+Ur/vrxeXnv1p3y3s9vSsX1+5/Lk8OH5at6nG+fHpcHj17XDVf7r75XdN+Xz+4dlg9f1mPnMR7VY2Satw7L0xq89xjjQT3GhiP5VxC4agR+85vflH/+85+vutuV/iA3Tg+W5cyto7IRFWeZe1E+uPOgfFnf+PD2RfnoaD633j45Lk+aif4RMuz4sHysZECJfg7L+8+rif3q0UF58vO76sYHyIODk5kMKl8+LA8fvap7WGz3oXx+/7Cc33pbPj5+UL4AyW39zgjhw8cX98tbBw/LSkx8KF88OABPtcxAvwcPq34/PL9fHtx7XipR9/F1+Riy+Lge16tHd8rb9+f3Hh0eL8i2prcPSh6dNqIL8ugIsvz2gxdKtoIo+j5S8vEDeDoAvYoLlmN3ygc8KNR5+fBgQW4/LU8amfoBz+GoGQfTeLBAo+FB/hUESqXXKB3i7t+V9N/+afnPX/+tuve5ZT8taemUgt//ZVkr/rdgy2pJIxsrusi3KOhgNdYP5qsxpqKZ1MEqTBUTq1dKsX5DSaYUYcXntqpbBqxVRgzXafWVrN6A2vW9+tKGf1KKYh2Wqrqy2aGe1VjxNjSpL2twzQ6wIlYeDaR7CiJYvuqmJlaqlM9Xn5spxRQmNnWc2i9iuuQSu03qFnAFd+yKaAsraD3L6k0VmynKHUHg20cgoUnwx+Vh/uX3NNlmTcun1OONA6ZBhtWl2PFooOZWRlFYULttVPQgc9qYyHFj0V7oxR7FsFTV9XSLHCOnrDZ8b8ZcJ8c1KJrWVnzIyAjxcY62e79MW7dhRVNiAq5ax6AsPStfkjAhq9smJc00yBW4dOcFstZt7sGSZuZ0lgR4RSdRLYDyCNa2e/fIwL8V9yyvuC24wGaFeNLEwrVgTWsRWydhVwR2rQW53aFOI1Mht2PIuEbc6W2XLMQHijFt4THJxytDYD9zd3Icxm1g/J8LON8yyDxHYWo5A5omLk36HZjBE5rClVnNa530Wn+B1FFEC/4LkrHguK6mG1TSoRSxiOBr+rxRU2PDvyxksQtrRgg0d43v4D5nVOFyKSIajyZQ+qBIpoiBg4J1bilyyuEWnsen6dRqIYaNB8KY6QtjRG8VAudSlQqCwDeOQItM4xbRHxcXhLfJMLYIGrVxYEJqVsLNOIArrg+lbtIuoFykcJ22KVQTGt+TlLSBkjRLRcNETadw3wURFJwMsbNE3dVKa763HOzA9EOoL1hoQeGBnxSzefd+maS+GODWuGlX+srB8mK9VVk2l4uQJPj/7AgJISoONm9FlPcNisOUnN6YWoO+WjjaCN+IIdeGShABrHhC3gTXoO2lkHs04OsYF4QXdLa6LMhwMJiHQyjEfiXLCmCY22v5aFrLv4LAZSGwp0qaS553n6J/+Bf6D4XsbToZedTdJDshHJMCApcVI7aa+VPKHex+RNwCdp1fUSmoYGmlBM86sbUDG+kYgmdMFsY6wI4uPRmQ7e/QTqoIAoLAJyDQoq43oonzj/Sv9YLw4D7HPu24jNEs6mGeduKMlNbG38eITTW2sVJQ2Lepn/XIGw7VwjPodNbHrq2SMRB/VQwQ79qj1jRHPBh3lO7Y7yqx3b/P5VrdZhd4cOyRnU4Q82Zh4cxWRYtaTkYeAoe7WIhq7a5anOYB4nFHOg29Pg2hHPNRIEpH47IkRmv5Wt/S+egT1JUiCFw3Avvp7gTqRhcuy3ev6MXzF/TqbQrBZm20AGUBB9xGC88KCpupIUj2HGUJq0oNK1k2OKkCM3u+ZHVaILn1IyxXGtrOCVGWzfsGJ/PWLPE2sJWHE0ogfEauhZ1WbGGDYNrab32TrWhsTZtVhmUvA09qRS9FEBAENiGgWX0K07f06sVzevH6HSVQGoxNlddc500DhZr4Gv7LltyWaTAiP1qdwXB1TlvU93vkwGXKc7RQWtAuBZsMoOhMpyEFqa3chbv3uwv9qg47EPIF/2sOGXnxwvzh+BB2cWpwy8J6aLs2JWFEIXZnOfVW/AgWNGvgIRzDVDJvLkRZfrEca3oGts3nGvMZalD6/EHA6qoUQeDKEdhbJY2R1rFydHkXkbld22jZLuId/GonJDfE7qkJfA6Ovb0dYTeXjV2R03ryZ9MJpdjRpGTf2sKuhXU3DMRS5BC+aXUTK8gJC2clRVpwPWYU1rEpOfoIN8hkjYVUklTxcvh74k2Jdb159U39I67FjDDmmrkkoICw88lYx6tcEwQEgSUEdBPxXohNtY2NC8FNiCklTd1EPJUNq9g4qr5j1+hkhJ2KZyiy8oF4q3oLYwG5NYasWFxP8uJyg4jA4bkWxd4QMVltOD0v0u+mEZy9bsHqFY9rpQcxeP7CbviztTddgVLmIC5tFFDRdlTUhQY5bUYezq9keVW1Y1dq2gS1cV+TtPJIQFW2bcjASbXzNY/G8x3+kNEWdpiiqir51CMP8lXbxIpcFwQuEYH9dHdeFFA+xmKY4mw0E6ZyXtEiKHY4IRzyvb1ghTfysD0cKzwfM7zQHfJ9xH2sbWWQ09GwOcCkKeIrwgEk8kKxhx6CbWHRC2DVYuUSLpNKDmvUHg5o3MGWcx9nEfG2diNd24PWHtIQ2/Db9gSrSgh9BMSaAeLmUFuD0Frun8VeU+C28UcU9bBiHfGPBgJufb8W4mu7kouCgCDwBRDQYHaqLGlE1nBMvU4fygVLoIJ0d0zjZTGBHnGMxAgbe3pYRLUQ0oAAeIgGtRhTBYqM0+/i+AvIpun4jFVPcyATki5i3eaEd+t398FqiHXzox6sW4j5amHjACz7ye7NZzVbiEvL+x5kcaORQfbpXRq0ejPZZMOKZmEhbgewKuIoJbdjI76MF5uQf8Cz0+tjsY0Fu4UND06tEOs4Dglyuwu+fFjoClBjubkoET+BXWkiCHwSAt/x9tBParljo9PTU/rpp58IW1Xp7t27O7aSaoKAICAIfDoCP/74I/3yyy/0/ffffzoRaXk5CCAQP0OMbxMukXo2dbWAwt7VqkFFBqsaNpFVG6JynHvpUIKdsaPKhHg5Yxeq3xQCv/76K/3www+EIziI/qa9PLY//I7oT2/5DI7PGvNeuzs/CzlpLAgIAoKAIHBhBApsuLJ7QR2rm+NokRz5Ss8JHblwL+c3iIY4OqjJw4e4szA1yDLObyc1BIGrREDcnVeJtvQlCAgCgsCeI6DCLrAZy7ZGsKYhfMTyaAxX41UXZ4idt/AFWwgT0RH7ZyKkxL16XfGqhy39fWUIiJL2lT0wYVcQEAQEga8aARxj1MUGiO51DwIHffuhe91cSP+CwFYExN25FR65KQgIAoKAICAICAKCwPUgIEra9eAuvQoCgoAgIAgIAoKAILAVAVHStsIjNwUBQUAQEAQEAUFAELgeBERJ2wV3nOljIFMAJyS3LBPZBnA4ZR8HwZ7bNiXPdshLz624UiFC6heD+tFF262vHyLZcme6/t60g5yg5zKIHII4V2lcn2W7npJcFQQEgc9DIKAOsnsYnGAdsobljMXpiS5p3nGKJA1p4i6J/EYoCuTSM7+QcNtNfm1i5VPl8yZ6cl0Q+PIIyMaBXTG1cZjrtMoHR9iu3be6NIoSKGG7EtheL510cXAi8vq1eZcTDk+MoKFdwU4jx0dqlTNHWm7nVe4KAoLAJSGwmGAd6tO0Y9Fg3KFoe8LOT2JG7yBjQZvTw329ReTX1/vshPPdEBBL2m44LdfSkDcOaZrShfyZn0JmsU0GpazO5ILLOFEcGQGqQxY/l/L29hoOc5QcnNsxkruCwPUggMwmyECSJenldK8hQwnSJn3NReTX1/z0hPddENhfJQ05NftI8Kaze8HpzfO27YIaVricWYTz6nEpkEuz10YaFqQpcZBWpb8hF10R+yrRr2XDlQH34UD5MZCTs+9QD4niwgHSLvHqFql8Z27SqE+m4y+4VhMaIbXLBE137VfxiFx0fc5TCjdKexDOXBwRclu5kzpnTDalAXL3WXab3C6sei5SzMxcrgUlkwFSpeBcIaSe6q0bI/O66D6JB2RqFo1mOV8KCmAZUJ4O5D8dcV+WTRbcO65X5dBjXnPk33PhVrbbOGxy5FHX7CBXaFUyHITZRhJDhXWbn1t9Q/4RBG4iAsixOe7aCJfAwgv5MAdN/t0dec0haBo5g0lMfhfp35ScQWq40XweqzkDuWI7C3OGpzWs/2Y3nPXGp/s7fh2ogXtGc4/rdZDns4f53Z0o+bBxrjEfSK9kYR7anF4qWucwZTnl0MDvI9UdeMbn7ng+WYs8Iq+HVEzg2UKaqJh5PSPrOMyiknXMC8sum8cI/tJ6REvyi2U6+nKQ58mCTG9E1Hq5e84DyMfgrYfE6pwuisfaIV8xSRQPTGp7EyUrXZ85wYG8ow7SS1XPpQ0c66pE+G3o8hiBleX0Z3mcN8ruzMc40e+oC9y43y7NYNswvm3vxTmjlNtfAwKcFuoyy29/+1vOiVAiLdRldnNB2m/LJ0e3FF+zP4en5euPG8i8flQe3HtWflC3P5bvXj0ujw8flC/Uhffls3vH5WnT+P2z8t7Bw/KlovWufHp8Uj59x5/flo+PjsrHb6pOPr4+LQ+PnuBqVV6fHpb3XzQMLLT7+Kp8eHC/nN16/3N5797P6HVbvzXR+p9Xj+6Ut+8/RwuUj6/LR4fHNU9lqfp9zv1+LF8+PCiPn9YcvX1antw+KE/fcKM35ePD2wv3nmD8j3F1pXx8WT44eFCPHcg8Oylv0a3y5JnqGQV9H1XYvHp0UJ78/K66/IHbnZRVNe7rcIbn++f3yzu30YZrfnhRPuD21YPA1wflAcZVf61oyd+CABD4zW9+U/75z3++Ziw+Yn4dLMuZW0dlM8XOMof3+w7mT33jw9sXmC/zufD2yXF5UgkTTFfIsONDyBOuzPLssHz0qpIfPC/u3KnnzMuH5eHDV7Ou3j0FjZ/r+Yh7B809fL5z+6R88no2uTbOtY+YkwcPX0JioLx5XB6dsDxaLZWM5XqVmHxe3r9TzfGPL+6Xt1hGqhsfyhcPDiCD+AvLpvul+sjl3dPyWMk6yISjRo5+KJ/f57FWVebyC3TuM52Kk/fP7pWHqtI2ubsonyt6s78/PCtPbt2e0fvAWNXy+s3jw/L20Wn5sh40y/KjRr4CFebpuH5OLOeqsUEe/nxSHj1h+bpFdkO+n9xakH8YR4X1pvFhhBvfi5UxydcvjgDrNUqHuPt3Jf23f1r+89d/q+59btlPS1o6peD3f1nWof8t2G6ViQZYNcIChoB+052SORpVp1MjPm2aYkWFpMeqtLDCNRJK6sXqvBMDq+iQhlZVT4M1yszT84N24Vp17YSCqFrF5eGUCliYWjv3yxxoyKuMNuojLGkmu2rnnFWfUopjJE6H9UwVWK86Na/qOw6gnN/DCg9r2TMkmFcroalaMBcUw2pg30Py5GltJUtCio02MVQ2cuQF2NCgim6RY+SUKcNiTDG1sfqscGq5OJm8hhaEKEbCaKcOotGRIN5K4CauqMjfgsANQyChSfDHZZ7+8nuaYF5sLPmUerxxwDTIQNxr7Hg0UBMgQ/qkgtpto2qK+cgJweMYkwZWqbBoU68+tV9vw6rTzJmNHZ29ocHS07frybVtriHpOyHvpZr/1oAC390Q14YNS/12da8FC58Fb0E9WXVYphAOhwKXrmOAHFOz4GmArMM4uWSQdRrLOtRpaRklKV/XYb0KYGFTVRZKSlFiABMl5ajljsjrmvj0iXIXLTUdXpFORU93gGkWzSxkVm9A7eoWpWFILVjwqq8sa23KEK/MhTMZ8NiY8xa8JJOOAdF4zm8GZGSnkX+wwukZt980vi3vheJAyteOwH5uHOA4rNt4dP+58PhuQfmqJ93ah7qwcaBIfAiDHo3tKXX1gnJ2E8LNUMWQYTolObk8K5cK3B16RqHvQ+GKoSQlFJG9WmnNd40c10AaFUx6uGejaUa2CiKOduy3IqmzYFUF0hv/n2EPV3KIGWuGAQKKmyaq3eL36hfgLA0WuDpO8U4hb3MIZD5ZvA3XyBSKVJuMKFS7ySrZrFE6hUs1iIBFRjGG1+Vu4N4pWsZc6EPIzWLzcvAYDvFD5fMQMAi0y+01Y+GbUgSB60YAO8KNW0R/XFwQ3ibD2CJoFjcOwK04QOhBH0rdpI13H4u6cbdNoZpA+J6kpA0wCzFn8pZZKwm4hRCOT4ln1TDhZ7rdlrnGaZ087NB0rAjuTI969obxqJi35hmAJ/5cCw19MRC2DhthwWQjHKI3iahwLQqDnOyhoQh0vR51ulhgQvnyhy4ZM0Yb+sCgWJAVWPjBW6jKp8ldNOS44IY8xqJD1hcN/wvCsQBWMTZ3tINaLkK254arhmoN4L7stMmCK9eH6xf6KDDY8pvBJADUXPQ2z2TT+LDQ3/ReNLzLv181AnuqpLnkefcp+od/of9Qj+82nXDs0wZZs/qENViZesg3F6W4gxWd1kK8Qjjij1tKTCOsxsKOR6MhYkH0CMJtuqX+/JaOODfdm1KKyT1NbaxIcQ8SYLd+azpnhNqarmsBVN2ZC6Q1NTdeMhC/kQ+hQHYQp4EfnB5iVhKti+TFBZlhQfaAQS4o7NvUz3rkDYdKOQ46nY1WxUW2dD6SAIqfFEHg5iPQgnKB/JDOP9K/1gvCg/sc67nLZMToYPXudWGNijPCGqf6Pp7ieJ6VkasdRzvSvABom+eaAWtWTHYIBQ2KU4AY1yn4/BJFg4XNHASwWFWyrlcLVd1GPFfcpkm/i9ivkMboG+vBlQIMzsDw6XJ3lfq27/YQ3pnOmc6VZ2UUxsCrD3lvUzcICWHAm2U3HvXmsm58qM3vybr3YjMhufMVIbCf7k48IKMbUPLuFb14/oJevYUZvm+dnd8bHySSAmuNEoPPsOjM5xYsSJ43CxCdkYD7IICtyIP538LyUp+139jJ/AYmOtbPNIUlit2Fldzasd8dyFdV4FLQ4XKcDQRj2iowNhCGUmanIUXgtUAgLeyT5DrABLtXpylcNBBQCL2lYNqivt+DcRBWTQjbYr5EJS3P5gobf260NKy4C7a0NV3DbeBDoKcbWJHLgsB1I6BZfSxQ3tKrF8/pxet3lAQdzIndC28a4HeetQ84zqqQgLp5GsCqxGEQbImCNWU2XQu2utSVcGtmvuJPiyuebWxsmWtZNMGYWP/oYx73KfcQKL+OFuSi8mKqwput2JuwruLCNd2hjhFRMIZ8xqYBJeuw+WIaxJRzyIUfkm8G5GHBt1xYJlcbutR1uCYDZvJz5C4EICNfFZZDLCNnF2YfNJgts0VhmWBTwZgRySkO4EVAO7s3pulQJ8+PcP1TZPeG8W17L86yKle+QgT2VknjZ6UbiB/jHY/mmpm39WHyhKkFHlYxjoGVHUzzqmQBjbBTccFeXV1nEzbM4Kmqhp2SYx9xJHPxqejNJOtq5xx/ktEYCkkL8Q5K7u7a7yqpjd8N7JzKaDKp4sfyCPETaoV+0WJDEYshjFIcxKlELP61KR2PKEZciq2usLBDDFxNv4iwqww/Nmr4OnafaVN8Z6AYpzHV4Xgg5JCFXarYCKtKPvXIi7H7rfoqfwsCNxMBHYdfux3ElhoXfleVkqZG1cL8hMUZicnVdygukxHvBMfb30KsLOTOJK5q5mEwPwAXblAtjqu4TbQJoOzsVLbMtQI7SQdBOifDrsC1RNM5vwgJmSD8AREb5xQdWGk0hmXKaGQdlmHjAc6pbMQRMOFYr+Vikm2mNA0zdTmbDqnPguIcubuNmQLxgf60kut5OKEIvxfr2DcQN5ZNJnVsbEGRzwcQV/zFsKDB6KgKP0s+TeDTZPeG8W17L7YNTu59NQjsp7vzCzweXhHmSqswqOsPKcLWeMtD/ADiImwf27dXZQhcoqM+bGnYOu8ZOPYDQe9trRa4oGLiew7XgRUMEGzcPsOhUnS8BCbzRhzu2O8ZSpsv2MMxdXqIm8CPgW7x1vPmB2Jzm7N3MH4MvtNv0QR0uGhw1xpwZxbeuP6RwtEeI5fcHhQyjj/DZoCOA8GqhLAJCxtcAziOxM517F/AsSR6UhHSO3BThzgGxMLBv8wbDv0dd+exOFUt+VsQ+GYQ4DixypKGNQrmZ6/Th7KGH3u8/bo7RlwsDxUuNR8LTmz6sbHpwISssFv1nDFx6DbCLDrWlFoGFnuwbq+KprVgbZtr3RHirBBIP2XeCMfnTNZbB3khqfmwpA9UvJg9wmeIr0bXWtsvLrbAr1EEaFdzqrVpOMDYsYnAg3WqMIZ0NuJBpw5cyyGO9XB8LPiwHPQnDogVW+XuJh74ugalzIDl0xnBa4IRDnyEqeB6rXPNmmrY3DFWGAeVh8TgIzQqda4zwlEgONakjXi2vLBpNLFUu51+M2Y98IcN48Odze/FEgH58pUi8B1vD71M3k9PT+mnn34ibFWlu3fvXmZXQvszESiwY6vApooq6DiHIHQowS7MUSVXPpP6rs0hzGBu1OASrkT0lDpwb/SwWcPZlYTU23sEfvzxR/rll1/o+++/30MssJAxseBKsFi8ttHzOWk90hGD1WtdkIkUc71LNAl717cA43PSHIwhPi/W+IJjk+rfFAK//vor/fDDD4QjOIj+ZmW2/eF3RH96y2dwfNaY99rd+VnIfYONoyG2fk+yamSI9wpTA/FzVz3QlHwXGztqQ0ARh5TAnXPlbFz1sKU/QUAQUAhkcNcWsyMtBBRBYL8REHfnfj//pdE7Q+xEg9/R8nljg0bm0KvOgrtSjOAK9RA83DEpQDyJhrV0zxuKknalz0A6EwSuCwFsMkKqAGdYuQuviwvpVxC4KQiIu/OmPAnhQxAQBL4YAvvt7vxiMAohQUAQ2IKAuDu3gCO3BAFBQBAQBAQBQUAQ+JYRkJi0b/npytgEAUFAEBAEBAFB4KtFQJS0r/bRCeOCgCAgCAgCgoAg8C0jIEraLk836iGNE/LwIcG6ZZlkmjiUsT9dyDKwC5Hd60w7Os5cS3dv8IVqhkh4jjSbX6BESF1jUD/6RFK8/R2Jm1fPI/pEatJMEPhKEECuTvM7+s4c1AejNmwjR3DLRcaSm1dyHFjG+XhxXNr2MsUZY92wqpPgWAucw/YlRE2G8+Hsi8jKTbJliaeUPJwnNyO7yPv2UcpdQeCLIyC7O3eFdCHBOuF4ir6FQyKjBJN5VwK713NwRDXO2969wY2riUNmkQZqwzHkN45bYUgQuDkIHNAhTWiEE/PHZ07EvjlcNpzoHWQ9aCN5+kVYM3vIw3nDxMNN5OkimErdbxYBsaR9yqPVkPrIypGX7ryzsz+FOE66xoGy63LEfRq162iFE9FbfIzHdfQtfQoCXzMCSMredyhEyqf6xMKbPRikhDJaF53oSAuHrCsXbXW5QNxEni53xEL960Bgf5W0NKA+EslxLjXD6V0wT2WVyJdzsXEp0glSliDFkePgj0t9nPNTlYJipC1yLNxrd2jg9ZAyqRK+8QCJxyeNksen+1s0qA9wjfrze1yvjbQrA9x3kQ+Tk/ZGI6QqgTme+2t3kNy4IQM+usgjZyHJn4Xkx9N1Up7N/TgJ3B9w2ifUszvILTdXNgvkxuxzPlO4dtuDULkxUpgLrVHNHA+MTwRHKpRU8YKE7xifjT8uLIAVpWV3QTYdIMUL+AVvNq+8FTicl7OLdriG6yZ4msGm7m8o0w6ZXeTs7CJdFNNzOU9eVfcMVkVMfhfuEPVckDZnVI2Ha2dIDM1j5L7t7mSWpJ2vt4GfepZtfi8q2kWA1FbcL9LOtBlj15vhvn58296LDWOTy98eAsiXOe7aCJfAwsVs02CanjNG5AVuIy1aznlpN1Rl2cWpn/Ce2khH5CN/7dqyVh5wJgAHKY765OLAWBufu+P53N70/nOu0DHPOcwL28I4mkm35ArMKICCWckDpKjqYq7PRcucxUWXo5Kd1fx0LJyLWLtON/KBmToBnqaSqV3yVI7fdWWTbFqoC/kwwFwecFLQTW7QdaTlmiBwlQhwWqjLLL/97W85J0KJtFCX2c0Fab8tnxzdUnzN/hyelq8/biDz+lF5cO9Z+UHd/li+e/W4PD58UL5QF96Xz+4dl6dN4/fPynsHD8uXTEt9bup9LF+fHpa37j1XdN48PizvP286/AAaR+Xjt1X/XK+5x/VuH52WL99X9z6+Pi2P7j9Hr1wqmsdP36lvrx4doF3F5fufT8qjJzXBqmn194dn5cmt26hXUfjw8mF5cPSk5JqvHt0pbze0P74uHx0el4r02yfl0fHTkhzFG6MAACAASURBVD9y+fDsXnnMtN//XJ4cPS7f8MWPr8qHjIka0rvy6fFJ1bZ8VT46aD5/LF8+PCwfcCVue/iwfFVD8PbpcXnw6HXTQXmvoVtdmf/98kF569ac3pvHR7N2q1i9fXJcntTYlB/xzI8Py8eK2Tfl46OGxofy+f3D8tErXP7wonxwhGdXQYivD8oD4MFfP764X97i56rufShfPGiw3jC+be/FfDTy6ZIQ+M1vflP++c9/viTqu5Ll+XmwLGduHZVP10zLiiLey8PqvXyPOXbw4EUtc16WD+/cL1+oSixvDsuHzUuKuXm8Ya6slweV7Dt4+LKi/f55ef/OSfmMxcGW95/lySHaVNP7KfqsZAYmdHnw8JXi7APmSDNfKj7vVHOdby7UgwBZM78Zq6PyHjOyhY+PmP/Mh5qGkFGPD2+VjfxTTDRlk2ya9f0RMvi4PIZAUGNa4mlRfuHeIu+LfcjnvUeA9RqlQ9z9u5L+2z8t//nrv1X3Prf8L1epEN6YvtIpBb//yzI7/xbMrCZr+YwGyupjIiDedKdkjkbVafyIT5umsHIh6bEqLVhsjIQSWLGKeEoJktAhZAMFCYZd55Ny0Vm9AbVbFfk0DKk1S5nCNG3KEBvHhbMEZGmmrFktWKwmHaNqtPK3psPa16kI6liJO1lUW4WYXrviEcmR2ya7dNEYFoB2MaVQWeYKCmGis2GFhF+WtDwlpNrEZyQPDoZIqLzSWRpTrGPVq1jRyBn41LPwsdWlIIJFrq7PmzIoX2f6W6GHrzosXNjjoOhZyJCgJfHMNTTHKqMoLKjdVhVRFRZJdBYrqwNcG1pGScpI6bAABjRgnpIpxUj27qjnxf24ZCXRLIhbh9WxepY6xm8orGnT+La8FxV1+fvbRyChSfDH5WH+5fc0OdeaxvMXcz4a0ThdRSmmMLGp07ykeF9diig6U2+bPMAGoX67iiNrIRWclYIm+tny/seQMSbkjpquRoc8WNDraTJjUIcVLJ506ustyMsW5TnPtx1Khji80KYBy6UtfCRg1IIcVX1D5nS7kEPryjmyqYiG1Atd8ofWDXO7rhuMXNtnBPZz4wDHfN3GY//PhUd/C8pXrQitfSEWNg4USPbdhot0bGPXlY6E4NmUBg6ElpJgBaVJTi5+/4ssR2yWMRcCCDT7lDgtXZ+LwyJnFyqUhaDSboosodxwlWJmQQFyOnA/Rn3y4c6oFKM1o+F4seYyYkp0jKFgAijzvkAf/1eXoYQ6KYRoTt0ONjXwj4TNtZFjcxiQC9fDFDE0HlI5VVwt9FlAaeQ+6kuaCddG/VmHIjMeTSiKUyg8McbRXmi4+aOm0kU1RAzSwXzN/gL/eC5QINlFE6rO8T1JSRtwTYO6cD13ulBE4S71h3BlMkFgm4dDKHZ+RR+8x7k9p70YKFi7uvGQ148PPG16LzaPTO58WwhgR7hxi+iPiwvC22QY2wRNjYDmUL9XkOtFSIu2gEqBUAsoIHM5ggVHKyeInDNlozxQcWRNddDizzwt+J1d+/5jfoG+PusUClizalzsFXxBSyNvElKMBUyKeU2DM2ytvRB7HuXdoFq0bZmHBeKAtQUhyrGv2pqxExaiG2UT5M6QsZ1EZJ0RWGvZk4uCwLUhsKdKGiawd5+if/gX+g8F/W06GXnY5r7bc9DMDqxBXrV6taDLtBAXEmJb+UpzGKjYUPPFiz2EJbCzRrrAijcKY1iGENOCeJVuEGIX6pp62zjaUJ1Xr4kfUW5CAMOyVv1uIL8n4rliJ4DQg3I49Sgcu8vWQhb+62imiI3DytsC7oMuFK0Elkp/G2OfcA/WwN54iuNAzrbVbcTsxW2a9BEzaIc0Dj2lPOptxLjhWIHVwsNYWzaND5U3vRdr6cjFbxABbALwkA/X+Uf613pBeHDfp5G7bkKcHb4JC7ppjSgY9M7e3OXKWnmwveH697+ojv84h+086JA90mno9WkILZCP59hJR8snNAxMxIYZM+bW80EUQptsFpSq8saJuU02Ie4Pmp3a+GWeM6jtcMldQeDSEdhPdydgNbBqS969ohfPX9CrtzD39y9i9uZJ3ggLfIY1JZs9KiQIxqqQg/bZ2JKx37MpMP1jkVgXFg4bJcy8zconXkVm2QLNBJsKxjFq5RQHcK8WcMX1xjQd6uRBqVpb0H6++MRntDlvN6lmu2QncAlj9dmCwqZEWxrSJMpIgyVvFEzgnvHOBv+z1QtuzGakOVzAU0QT5+EEW/dH+MGysKrnLfxza9hanhcu5gv0CNayHJbGs7owngueCoyZs5IGsJpxkDCCoKcBLHdwgXb8kHwzIC/EdTywAs9o9lSw4vYHwWxTwVq+NoyPNdNN78VaOnLxm0RAs/oUpm/p1Yvn9OL1O0qgyBi7jpTDEtoxeQjsn4sNvOtsTZtdyCEP1s3fLfIA8kqFMajCm6CwqYEn0Mb3n+Udas46hbyEHFiQQopSBAuaNfBgZTfVnEaDqotz/k78EaVdWLCbSbyRj8qaly9M6hwDWStFt8kmLN6GfpfiHo4C2Y3Fc0YgtwWBy0Ngb5U0hlQ3ED/GOxnNsz/x2yGvhJZa0WHCOwYsMUE927OARrwtCyR10yI9hMKkbhVKiDWCrQXfagKrlxIw2K01YeVhh2JgN1I2gYKj6oImBFwIIcslhgUNGyxV4Z2nvHN1XSnyKfm1dGJlKQIO5rqKi9fgfnHNKQ3HcHI6NV2MdQirnhoT+tOW3DB14xZ2QuYh4vb4O/gdYWcpmNdYiCdJjUdGEw90AMEuKBTAdKI6hesXmGpwt54dKVwyNlEwjiqaUMwmcMkmSr2EG3QwohnkCitct7AzDbtbJ4pX/MbAMughhq1Ct7p25u8N49v2XpyhIRe+bQR0HH7tdhC3amx/l86gwDGcfYLZn8LZxIC8MSOaNLsrsXAKqIn5XCawWR6k83mBUI1JAv5YAGx5/01UiGGZb+TVYBQsLPSqfnUcxZE22h/LGEykJavXmfHhAuoNxwYNegsSaCsfFhTdadU3dmeOp5UkPEP6PNkE5XncS+FSPjuOM7TkgiBwjQjsp7vzCwDOK89qZYn4Jn9IURc/8B4C6QsE3/tw5fEvO07N96EEdhEUb/CRFjjywkirzlvukLoTHMnhID4EsVgW/j2raJxlVMPRF2ME+3esAIoFrE8GlJ5RJeA6UICCHo7lQIxZXiCQf2KdJYArGpQyAyt6Z8TWKwhIrCoNXK9ZW9sGrfBDY+G4EcSUteoq9oBGCCJ22xNYy3LSez7BgLdcoNwNfYd6wCeABaAw4f6Bu0ejIYQzNlXYE6y6OcbFJROKLv8IbFWKcL/FPxgYp4OVfYENCL6/IOAXeregUfY6fShrUMJAWXfHiCPkCm0aDnCvzW5b9GnwwaF8HQHRXkhdjNOHxbKAA5t55OHOfiMX6KuPG8an4t42vRerNOS7ILAJAaNLfXtI/33aVIALFQuzCOEFzojf0Raswf6ZUAteJa6XB1BqeGGp+VhsDSAnIK9G+Kzm7eb3H35buCOxmcEZwdqsIVZucmZhZ8OKZkHe2QFifls4RgiBq7GypjUC4+wgi9CH9R2LNcTSjvk2DpUde1v4QAzpaNoD72PE0nEfDrwW3MdK2SSbFqqafWxiQixxd2JToOa/FEHg5iHwHW8PvUy2Tk9P6aeffiJsVaW7d+9eZlc3n3Y8IAu7iUIoSKu6zJUxz+cBOTgrKT4bQ3ceD/nEpTbaRbVSeF79S7mPc9KMoIej2pxLIS9Evw0EfvzxR/rll1/o+++//zYG9MVGweek9UiHVay3WXf6Yr0JIUHgW0bg119/pR9++IFwBAfR36xo+n/4HdGf3vIZHJ8FwV67Oz8Lub1rjKM3ghRHdKy3Wu0dHDJgQUAQEAQEAUHgkhEQJe2SAf5myBchBQmO4ljvQf1mhikDEQQEAUFAEBAEbgoCEpN2lU8C2+nj4Co7XNOX3lXJjS9cNMSM4MiKay+If0tvABvXjoMwIAh8EgJ81EX4SS2lkSAgCFw9AmJJu3rMpUdBQBAQBAQBQUAQEATORUCUtHMhkgqCgCAgCAgCgoAgIAhcPQKipF095tKjICAICAKCgCAgCAgC5yIgStq5EEkFQUAQEAQEAUFAEBAErh4BUdJ2wTz1yP7uOzJ69YnbTZu4T6bjn0mPsgvJy64z7eg4XDc9t5uojwMhJ9VRrZxrT0M+zTVHQ55LZ7lCgQNrTepfZIMCn38GfFfLMk8pebZDzbAWeV9tJ98FAUFAEBAEBIGvHQHZ3bnrE7xzSNoUp+Vnzs6J2HclfRn1HOSHmu6eJVCxoHeQNgkJ9K7toN01QNxEntawKZcEAUFAEBAEBIEvjoBY0naEVNPb1HdhyUGy46+haDpSs1xU29KQogr5925UuYk83SiAhBlBQBAQBASBbxWB/VXSkNS8jxyQnITccHpIMrwxO2P97JHjrt8nHXnzphuqZuEASZRt5IrEH+SYa3Igr7482bRPDnJ52kiWbndx7hdX4HRNSNfiDzoqx6dld5AsvekoRyom5NqEq89xkJuz49PsFo+jzdfRBuMIsqq3JVcgEpyP2kh0jkTkFpK+u0gAv3YIU+Tw7IaKQA4e2+iL+7ONueuUeW9jfHzdaTNuVX+ERM2MpwUe2x0khN7oM+WxIFcpeLHxx4XF7wwvqQ98cJ4b01jgqe5J/hEEBAFBQBAQBPYCgT1V0pC/zu3Q//2v/0b/+Zf/pD/+6/+k/+4OKDqjLay8A0h2PHCg8ExqTWjxdjGlQS8mN4goiiLEZCE3Xn96VgGhmLwB14tRL6R+PqRRWBEq4glNLY+mYUThUKMRkhqzDlREQ+pFLk1QPwynNGx51PNT3MkpgOKYQtEL0WbaTWkwrIkt8BYOoPC4AcVxRHE0Is3rw227yPzZz3qb+eD+kHwZiYw7bQPdQSEc5NSfgj/cC3o5DQYTFcMWjQaUQMGLweMUvMczLXKFdjahARK/j8FLBF50b1XpTcjr+mR6HhLFn+VLrggCgoAgIAgIAvuCwH4qaemUgt//ZfkZ/1swtwptfPoatfsdykceVK2VkkwpMjpQaKrrRhuB8HGolKzlolNLyyhJWSPUYUkKaGBVNTTdpX6nIqA7sJxlUKpQLYVC1IK1rLoDi55rUxYx5ZSixIDFq7rTgvXO657NrWmPYihURtWJbpFj5JRttHQtc1sgDm9sDKjHZDHG2HTJqZUnve2SlUQYI/iIMZZ2PXizQz1rg9sUlkstTynl/jWbRsGQnFnVghKvR2PTI6/pZJkd+SYICAKCgCAgCOwNAvu5cYDjtW7jGf/nwnO+ZZBZ6xhbn77Vp77BysWAJsZCTWg9Ras1D7rXEYCfZ2t2ShrUhSLS6VrUhlLlD10yWElhpWWxPWKxdL2gAkpakRcUj+EGDSptpsgSyg0XVrqc8kKDy7bmAwpY2z7LvYYK6dQHzxGlWUYx9Lvu2WprrqTkQ8FzvXE1LvCRh0Nqt31SXRaglduKjyyHS3Rm+QLvm2LboIh6wwA5QG2ajsbkdcyKFtOLPepGFvmRM7+2hiu5JAgIAoKAICAI7AMC+2lJY0XBu093Zk/4Np3AOtbdRUmDPavTdyjyJnRuGNuGN0i3+8ifGVA3HyHOrL8lfmtOwB5OaTqt/oRxCs+jWysyUJc2GK2q1gWFfZuwcRNWwCE2PvjU32TlWuG3CKFEalBKa0sf39bbiLWr+ZiGMWUZ4sdUu0qhrEhAs9xYNDLhno0Rv6fjyA+rG8yPMIFiq0Hx29XKt7ELuSEICAKCgCAgCHwDCOynkoYHZ0A5SN69ohfPX9CrtykUGWu7rrPwsLU23H85XHKLQWzYSqnBSjXzIuawcukLlrWmfZHQFPFouYY4Lz8k3wzIC2ulZrE9VJe8gGsU1im2hGW4NysJ4rrGMb5qpGnop+kU7tEgTFdeS7g6py3q+z1yzGrHZ8HmuXNLRmPEtznYyDDTXTUNVj1YDJu2RYSNDgGcnezCXeCDLWvZhj7SELF1GWmwBI6CCbUjb7bZAQel0Rgxb4Pe5EaePXcuZFJBEBAEBAFBQBD4ggjsrZLGGOqGQ24HMVbmzE+3I7QmdfsGBd50rpSZbbKx03Ja61LZFLs2LYfMMxRTKCKj+SYFKD46/nAp8in5aksjvJ/hBDFuVXsDu0CzCSx3VS2KsMM0zLmNSbaZIsC/6jSbDqk/SVWteWFFL4WLs1KaisinMZRLeC63lgJB/diegI0SC2Y6jMeKx9R0kU+hqMY5VEUDOzWxiWGaVjTTCRQxdLCujyygIayCimOMW0OM2sxdi0tGd0xD/Ncd17S2cik3BQFBQBAQBASBbxeB/YxJ+wLPs9XBcRuD/53+Z0NLa9PIw45OBPX70GsK3SHfxwn+q32h3nAwpl67DSUIViljSPD6qZg0DUqZEWDDwAiuQyg+Ax9HYnB7x6MxNhJ0rAAKDe7B4uSPTEW5g92RYQ9tfI4MQ98TR12fF5N6I5fcHo7NaBmkI/C/gyqbDF1Nu3gMa1ZqgM+putTCsRrjbgdu4pC6LuLGoFkVZNFw3FWWttbQg+uyS04ApYuVX1db7w62BzTiDRbtCTYQ5KT3fBou6chwJ/tDmthd8m3guTAS+SgICAKCgCAgCOwTAt+VKJc54NPTU/rpp5/o3//93+nu3buX2dXXTZvPSXNwNEg8guojRRAQBD4HgR9//JF++eUX+v777z+HjLQVBAQBQWAjAr/++iv98MMPRHf/juhv2NqyUP7wO6I/vaXPVbH22t25EXm5IQgIAoKAICAICAKCwDUjIEraNT8A6V4QEAS+PAJsSfurv/qrL09YKAoCgoAgcIUISEzaFYK9tSsdWQF4w6YUQUAQ+GwE/vmf//mzaQgBQUAQEASuGwGxpF33E5D+BQFBQBAQBAQBQUAQWIOAKGlrQJFLgoAgIAgIAoKAICAIXDcCoqRd9xOQ/gUBQUAQEAQEAUFAEFiDgChpa0CRS4KAICAICAKCgCAgCFw3AqKkXfcTkP4FAUFAEBAEBAFBQBBYg4AoaWtAkUuCgCAgCAgCgoAgIAhcNwKipF33E5D+BQFBQBAQBAQBQUAQWIOAKGlrQJFLgoAgIAgIAoKAICAIXDcCl36Y7XfffafG+Pd///dyAvh1P23pXxAQBAQBQUAQEAS+CAL/9V//VdGp9ZwvQnSFyKUraU1y0devX18G/0JTEBAEBAFBQBAQBASB60OgLC+t7++gRF0e9R3YVpa2v/5bov/t/9yhtlQRBAQBQUAQEAQEAUHghiPwh98R/ektfa6KJTFpN/w5C3uCgCAgCAgCgoAgsJ8IiJK2n89dRi0ICAKCgCAgCAgCNxwBUdJu+AMS9gQBQUAQEAQEAUFgPxEQJW0/n7uMWhAQBAQBQUAQEARuOAKipN3wByTsCQKCgCAgCAgCgsB+IiBK2n4+dxm1ICAICAKCgCAgCNxwBG7GERw3HCRhTxAQBAQBQUAQEAQEgYsi8LlHcFz6YbbnDejk5OS8KnJfEBAEBAFBQBAQBASBvUPg2i1pe4e4DFgQEAQEAUFAEBAEBIEdEJCYtB1AkiqCgCAgCAgCgoAgIAhcNQKipF014tKfICAICAKCgCAgCAgCOyAgStoOIEkVQUAQEAQEAUFAEBAErhoBUdKuGnHpTxAQBAQBQUAQEAQEgR0QECVtB5CkiiAgCAgCgoAgIAgIAleNwBUqaQkNzO9I03TSdf7Tos7/9wt1WzZ56VUMO6Wg55DRQt/osz/Nt3RaUDzukWO2qAVeW6ZLXlxsqb/brWRkkX2Zgw17ZLQntDqyeGCS42e7MblUKyPfMakff0LTHZqknk2tzpQ+H9mznV2Idtwn0+xTdJZMfSUlz7ZocEk4bOxWbggCgoAgIAjsNQJXd05aHlGc3aNxNqWOvoD5//V/XMkDyKB09ZIOhWlIxdAk25/SqN0hbU3vedAld6STP02pbeQ0bhs08CPq+86a2rteyikOc7IGxq4NLlwvjWPSnT4twkuUUQztw3JbF6ZHRUxRapFrXrzpLi2MfgTuLqdciLblUZJs4YNxyEzqXhIOW3qWW4KAICAICAJ7jMDVWdLikGLDIWtZg7gy6IsWrGFel0zWymC6MQxjrYLGSs3Em5I5HEFB48otcgY+eV3rM3mNKUwscj6XzEYuYP2LMrIsY7kGFIwwhSXtU/pNY4phdbTWabIb+fgGbyT1u7vvOHyDj1aGJAgIAoLATUbgypS0JIKvyLJp0RgRD6C0+KnCp4g9ci2TTLtDHQeu0ACaVDGlruESfySKaYD2ylvI7imrQ722SVY/xAW4Mvttsh3QNyzqTrjScjHaPepA28hjnwZTg/p9e7VK9b2IaBqb5DpzbdJwutSxq+955IE/m2zwark+JYq3gpJxlxwbCo1pUnsUVS68IqFxF9csjNMdUqjj83lKKiyNfWhUFtxrptWjyivL7jYoi70eubYB92tzHYqZ54IX1IdV0ItYGVvRJFjRguI57jA2JnXGaTXOtf3gVhpQj/sH1m4frlP8a5xBqqCg08L4u9S24BK2BhTWOMR+p+KHcfDwzPm5mQb1wppIMgLtAcXFhFyjS3yZ3bGm26Nu24I7Gm7FSBEjykMawN1qWg51XfAxI1LTwj/JuKMwt9wOlGqHlFd3gfb5NNil27gy6/eIsTfnLvEM726eV8/FMNsz13eBd6njOGRjfCa/C3O25JMgIAgIAoKAIPD5CCCv1BWUD+Xz+7dKunW7vH2b/9wpH754Xz67d1ievkb3H1+Xjw6PyydvP5blh+flvVvH5dN3uP72SXl0/LTkj+WHZ+W9g0clV3//80l5+/hx+fpDxTp/P3z4suSvH5/fL2/ff16C0pny+tFBSXdOyidNwzM1cOH1aXlwcKr6eYX6it87D8sXTPDjK/B5Uv78jhu+KU8Pwf8b/vi4PDp6XL7hOu+elseo/1J9PC4PH71SfL0+PSjvPHzFDbeWt88elo+ev0ed9+XPJw0+L8uHd47Kx6+5gzfl48MjYIWP4PWw7vfjywfl7UPwsEL9w/N7wOppydX5862TZ4qftf2gz2f3Dsr7qv935dOjW/hcg7xEl8d+UD58yfeqz+o5fnhZPn7wtFRsvnpY3rnHffE4bpcPGBB8e/HgsKLJvKv7H9Dn7fLkKXPIze6U954x3aruvWfv8Bm8HN8q763ywu/HEZ4V+vv4Ev0dnlbjX6DN/W2l8fFl+eDgQfkSNN7xe1Q/L4Xbvefg4mP5EjwdYoAf689HCvy35ZOjg/LRK+78VfnkYT1uNQopgoAgIAgIAoLA5yNwRZa0hMLYoMdRDosE/4Flx8W1BLFhcMMVoU9TAzFjjS8S1hQYjCiPwrklB+7SBBYVCzaqOEKs2GBIlXEro2CC2KZoRJ12m9rDCBsD1rsybVjt3o80GvXHsL2tLxniuth6xN5BtvLlgUtaC9YcGKgUn7BidcCb8pkWOhmwjEXjgLRur3YL4jpvNoBdZTLOycVmBWU8w2VL+RxjGsEq1mq1yOgGZ4L8DdslMx5StwPLGDCz2fSIOLoYVqeezUxklBawHIEH7leHdU0Zz0BfW7FUcrdxmJDdg5u35kE3WsrNu7YfWNcmiUN9Fb9W4D88B5j+OEbPAL8tuD5HbBzLEF9ILiyZPDK2emEs3ERH/Fq7oEkfMX39KemmhbEDC1QrctRLxzRKuzRCUGIK65QGC6AOSlFiU1cFfOWUZjqsWGgAy9U4gjVLgc28sJVQITkr8XgM3PvEsHAdDe+HgquhvQMNgiszYTe8BivaOFt6XhpvMsGbwu9uD5bXuhv17HjMYJ8mXZcGISyz44aPRQ7lsyAgCAgCgoAg8OkIXI2SluJHvYDLT2kKdUlwrQVXEX75OOC9+sHGvVoZ46qsYFh1I3aXtuBO1KD88I/m3K2XqB/5URjSdDqF4gclxrcXOoIrMoBylVaXdPzAqh/bDSWNU9IX4tUW3bRJiMB8aJVKVcCYIoLb00BgPnZ+NrFgeRRRyspCzsHmzZhTimIddbglu/MyyjLwOXYrWg0vqU9tF4pHZ0jeAGOtY/hydrc1bsdZfBTT1KBEGao182bCDbs8thS8od/aVRtHjKdF2oZ+GPvYZIUFBMF/mEOJxoPQwVMKfjMoZ/BQwzVdjREfobBhnHi2llFQCIW0H1vUH0FhBlvVs9Og1OhUQLmcDsdkDHpQpCpF2+Rny+5YHe+BApXj9irFnZWnGMq60ttrLJXCOivYiBHncIWiI5QE74qBZ6Mt0j6XBpOOqYC21YKbW/Wt+igoUu8eRsgbXnIen+oFdfDuKWZ1ao/B78igsGNRZ5JxBSmCgCAgCAgCgsAXQ+BKlDT+UU+aH/+a9RzXcvw4GvheZGyN4ZJTgF2XlSIEBSOBvcLADyKC38f4EbSUeQ0/6kopqJsUsMyhuYYf8yLxyMYGgUle31P/wEIy7CNeiy+mNPFDMjoutbKEklm/C/VBJ0+zysJVpFD80toChuuwAlYFSoY/phwxWTbivVAdDFR9jf2Y2t02abAWZrXKxLFLYygAS0rqQpfNxzwMKLERmwVlLvYDyoAP6wyLyipb+tiE0wKHGbPD/SJ2y0PgnoX4sKXCmwYYQ66TBeSHiElro+WGfgrwDHuUIpFOfIr4mS1TVN+SEM9OfeJYPODQZhxiCkKdOogNNPKAxrAu2XV8XAtmNo7lG6UdGrp4nguKdgFrKY/TUITnijs/VOaFuWFeQlZYVZ9NYYssf8bDL0LyAyhs/H4sKvHn0oA1EgCZShkDvUbFrbHqAiveGoslRA2zT4HWUbtdIyjRjod3sjNCDCFUQ2Zj0zu1wLV8FAQEAUFAEBAEdkXgSo7gYCtPy+pDsZiXWFl+qp9dE0qTDveencBKBaXC6rFqgh9dmFGGCHjPsUEgTesjENjShuDt2Q+2ZlOvk1MPgfwjrUVtf7J8xAfUE8pwVQAAAx1JREFUnHavDRecSa0BrEquh88tmsKlN3JgBeobS1jZ/RG5CGI3Wz11ppthD2io3Hrgq9OFe61N5gRuMLhnJxPmoqBOzyC3AxekDqtRZ0yTNqsWbeqYIyhcDoLroSNho8NqTP9Sx/iio64xgMLTxmYA0G05Bq6ypVAnG7xzYYXNdKvRu1B4XBBPcERIDsW1PwOlpsxYsVWwC2UvhdVtNMG5dLi3th/oe9i04QzgtnMCWDhTMqA8Vb3W9NQ/bMFi56MPJWwIa1+XxmO24GX4nmHTgUMRsM2hdikXKIoOS1o8DMgN8Mz5gnKXYoMH60D1e6B0TaW49yuFDby4Bc6qc6aomcKKusqLAWXYwq5bm9pOC1Y+WCh5/Au06VwaCcaCDRlD9A6rbtfxgKFFQzx3h98j8JfCWqZjY8AQ7xeQp8HEU2PI8e7meK+sAOM0PfI6Ok3xDqx7pxQIUgQBQUAQEAQEgQsi8B2HtV2wzSVWj6hnINYsxq7Os9rBJfYrpHdHIMSOW4/acbCiDG+mUExgufSgvEX9SknbXHXNHd5JaiL2CzF+SvldU3jHaFejAPSNNbdZkV5LI59AgQ5pCPevvbadXBQEBAFBQBAQBK4PgStxd24dXuaTo8HSlbPXagy3lku14WprM7l5TQjAJZlgA8FKDP8WZrBJxE/IHXIs2q4loRE6QDIC5aYdR4j5qnYHzAgU2NCh2x5sbHA9jyfY5QE36xL5HWhEU7UZZXe+duVf6gkCgoAgIAgIAp+PwJW4O7eyiR981x5SHxHbQw0uq0l3yS26ta3cvHIE2CWZwSW5q2JTYNetRzjXbZMVbO0IEJwPF67XZRcyXL1w06pQtoWiWS452QBxfmO4I7vkj1c5OocG4hfdfopz49prXLprmZKLgoAgIAgIAoLAlSJww9ydVzp26UwQEAQEAUFAEBAEBIEbi8D1uztvLDTCmCAgCAgCgoAgIAgIAteHgChp14e99CwICAKCgCAgCAgCgsBGBERJ2wiN3BAEBAFBQBAQBAQBQeD6EBAl7fqwl54FAUFAEBAEBAFBQBDYiMD/Dw1Sk8Rx6cBzAAAAAElFTkSuQmCC" width="338" /><br /></span><li style="clear: left; float: left; margin-bottom: 1em; margin-left: 1em;"><span style="font-family: arial; font-size: medium;"> <br /></span></li></ul></ul></div><div><span style="font-family: arial; font-size: medium;"> </span></div><div><span style="font-family: arial; font-size: medium;"> </span><span style="font-family: arial; font-size: medium;"> </span></div><br /><div><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;"> </span></h2><h2 style="text-align: left;"><span style="font-family: arial; font-size: medium;">✨<span style="font-size: x-large;">"Society: Public policy, business, economics, health, societal implications and confounding factors"</span></span></h2></div><div><span face="" style="font-size: medium;">Here we had </span><span style="font-size: medium;">discussions around:</span></div><div><ul style="text-align: left;"><li><span style="font-size: medium;"> How might we be able to measure competency and available potential of a workforce in the wild? An issue that can emerge is that it is not straightforward to know what are the best ways to visualize competency and potential. It can depend on the strategy and the type of work a company wants to get done. </span></li><li><span style="font-size: medium;">Social Glue. Having workers keep social interactions with each other can help them to get work done (they need social glue.) However, we discussed whether this social glue was also needed within crowd work. The group argued that since crowd work had primarily micro labor, social glue was not needed. Workers were transient. However, based on my lab's research around crowd work, I would argue that social glue is very necessary in crowd work as well. Primarily because: crowd work also has a steep learning curve. Therefore, the workers that stay on the platform are the ones who have invested significant time in learning how to find good work, and learning best practices for getting that work done to ensure quality and high wages for themselves. Crowd workers are not transient. Considering that these workers will spend significant time on Amazon Mechanical Turk, and other crowd platforms, it can be important to think about how might we facilitate empowering workers within this space. I would argue that it is through the social glue. Through creating solidarity between workers. My <a href="https://dl.acm.org/doi/10.1145/3274306">lab's CSCW paper </a>explored this angle of helping workers keep social glue through peer coaching. <br /></span></li><li><div style="box-sizing: border-box; font-style: normal; font-variant-caps: normal; font-variant-ligatures: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px;"><span style="font-size: medium;">How might we design technology to make workers more productive and flexible, what kinds of policies should we create to ensure that a diverse set of workers benefit from productivity and flexibility? For example, could we use breaks and work disruptions as mechanisms to help workers relaxed or further connected with their children? What kind of social divisions are new technologies creating in workers? (Some workers can access highly specialized tools because they can afford them, while other cannot. What type of divisions and gaps does this cause? ) Alex Williams has an <a href="https://acw.io/pubs/cscw2019-tooling-practices.pdf">interesting paper </a>looking at tooling and how it can further create social divisions in crowd workers. </span></div><div style="box-sizing: border-box; font-style: normal; font-variant-caps: normal; font-variant-ligatures: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px;"><span style="font-size: medium;"> </span></div><div style="box-sizing: border-box; font-style: normal; font-variant-caps: normal; font-variant-ligatures: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px;"><span style="font-size: medium;"> </span></div><div style="box-sizing: border-box; font-style: normal; font-variant-caps: normal; font-variant-ligatures: normal; font-weight: 400; letter-spacing: normal; text-align: start; text-indent: 0px; text-transform: none; white-space: normal; word-spacing: 0px;"><span style="font-size: medium;">Overall, I greatly enjoyed having discussions around the Future of Work with Microsoft, other academics and different folks for industry. It was motivating to brainstorm together to start defining an empowering future of work 👊 <br /></span></div></li></ul></div><p></p>Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-16826455862256670192019-11-17T22:56:00.000-07:002019-11-19T00:26:52.476-07:00Artificial Intelligence and Work: AAAI 2019 Fall Symposium <div class="separator" style="clear: both; text-align: center;"><a href="https://d2jv9003bew7ag.cloudfront.net/uploads/Diego-Rivera-%E2%80%93-Detroit-Industry-Murals.-Rachele-Huennekens.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="https://d2jv9003bew7ag.cloudfront.net/uploads/Diego-Rivera-%E2%80%93-Detroit-Industry-Murals.-Rachele-Huennekens.jpg" width="400" height="267" data-original-width="800" data-original-height="533" /></a></div>
Last week I attended the AAAI 2019 Fall Symposium on Artificial Intelligence and Work. The purpose of the two and one-half day symposium in Washington, DC, USA was to collectively create an action plan on how Artificial Intelligence (AI) could contribute to the Future of Work. There has been several conversations about Artificial Intelligence and how it is displacing people from their jobs. Our goal was to shift these conversations and re-think how AI might be used to empower workers. Our workshop was organized by the awesome Larry Medsker and Kevin Crowston.<br><br><br>
For this purpose, we first had a series of talks from scholars all over the country who presented on how they are using AI for the Future of Work. We then had discussions on each of these talks, and created teams to collectively work on potential NSF proposals that addressed interesting angles of these talks. I (Saiph) personally gave a talk on how artificial intelligence could be used to help workers produce collective action and build the societies they envision. The slides from my talk are here: <a href="tinyurl.com/uosprgw">Slides.</a><br><br>
A few of the other talks included:<br>
*<b>Algorithmic Management and Algorithmic Competencies: Understanding and Appropriating Algorithms in Gig work.</b><br>
This talk was given by Prof Mohammad Hossein Jarrahi. The talk discussed how algorithms were managing gig workers and how this was leading workers to adopt creative ways of playing the algorithms to gain agency. For instance, Upwork does not allow workers to post their first names to control workers getting visibility outside the platform (e.g., a potential employer might search for them on Bing, find their website or email, and hence end up contacting and hiring them outside Upwork, which is a threat to Upwork as they would not get commissions.) To overcome this problem, workers use social media creatively to engage their potential employers outside the platforms and facilitate that they do find them. Prof Mohammad Hossein Jarrahi talked about the importance of facilitating tools for gig workers that could operate cross gig markets (a type of META-tools that could help workers to manage themselves as they work and operate in different gig markets.)<br><br>
*<b>Modeling the Future of Workforce Development in the Age of AI.</b><br>
This talk was given by Prof. Ellen Scully-Russ who discussed how Artificial Intelligence is creating situations where a lot of workers are being left behind. The problem is also that if you can't enter into a lower wage labor market your chances of entering into another career is even lower. Professor Scully-Russ questions then, how do people enter the workforce? The lower level jobs are being taken by machines. The workers who would normally join at the lower level and then scale their way up, now can no longer even join the workforce. What can we do about this situation? This problem is especially aggravated with people of color (Latinos and Blacks are the populations who have the most jobs that have the risk of being automated. When thinking about strategies to address the Future of Work, we need to consider workforce planning models that can take into account that certain entry jobs might be removed, and there is therefore a need to think about plans for inserting people in workforce levels that are higher than what they would normally enter. Another problem that Prof. Ellen Scully-Russ discussed is that workplaces across the world are integrating Artificial Intelligence into their operation. But, we currently do not have a clear idea of how this is affecting people's work, how it is affecting workers, or the overall effect of integrating such technology. We need to think about the best ways in which the community could share results and collectively derive design principles on AI and work.<br> <br>
We also had a great talk from NSF program director Dr. Stephanie Elizabeth August who provided an overview of how NSF views the Future of Work with Artificial Intelligence. NSF is investing 8.1 billion dollars on Artificial Intelligence research. In specific, NSF plans to invest in the creation of centers through which different institutions can work with each other (sharing code, research, data, infrastructure.) In this space, NSF also plans to support research in the domain of work. For this purpose, NSF has created the program of Future of Work at the Human-Technology Frontier: Core Research (FW-HTF). This program focuses on funding research proposals in the domain of work that would not be funded by other solicitations. Most importantly the solicitations for the Future of Work need to consider the three areas of: technlogy for work, the life of the worker, and the context of work. You need to have research questions for these three areas. <br><br><br><br>
Overall, I really enjoyed the symposium and felt excited about building and transforming the future of work.
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com3tag:blogger.com,1999:blog-30573458.post-6990714877170768152015-09-04T13:51:00.000-06:002015-09-04T13:58:01.613-06:00Tag Me Maybe? Maintaining Friendships and Influencing Algorithms<div class="separator" style="clear: both; text-align: center;">
<a href="https://cdn-images-1.medium.com/max/800/1*hsTcPbFzpwqrQvJk68zjUA.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="https://cdn-images-1.medium.com/max/800/1*hsTcPbFzpwqrQvJk68zjUA.png" /></a></div>
<div class="graf--p" name="41b2">
<br /></div>
<div class="graf--p" name="41b2">
<br /></div>
<div class="graf--p" name="41b2">
[Caption: Example of Facebook Post with People tagged.]<br />
<br /></div>
<div class="graf--p" name="41b2">
With the “@” sign, we are now able to include others in our lives more easily. Tagging or mentioning people in content lets you share the story with someone in front of everyone, demonstrating publicly both your friendship and the shared interests that created it. We did interviews with 120 participants to uncover perceptions on tagging people.</div>
<br />
<h3 class="graf--h4" name="1c70">
Tagging To Build <b class="markup--strong markup--h4-strong">Stronger Relationships.</b></h3>
<div class="graf--p graf--empty" name="ae55">
<br /></div>
<div class="graf--p" name="41fd">
Most (79%) saw tagging as a way to build strong relationships, citing: <i class="markup--em markup--p-em">“it made people feel special to have someone making time for them to tag them.” </i>Tagged posts generated automatically by companies were not perceived as positively, because people did not connect them with genuine human connections. As designers we need to help companies better engage with their online clients, one option being encouraging more humanized interactions. The discussion matters given the recent lawsuit over people’s names appearing in tagged ads.</div>
<br />
<h3 class="graf--h4" name="80ad">
Tagging to influence Facebook’s algorithms, and distribute more <b class="markup--strong markup--h4-strong">surprising content</b>.</h3>
<div class="graf--p graf--empty" name="9981">
<br /></div>
<div class="graf--p" name="8a22">
Our interviewees (61%) saw tagging as a way to expose audiences to information outside of what they saw on Facebook. There was a notion that Facebook showed to friends a same type of content. Tagging broke this by letting people reach new social graphs at once (e.g., the friends of the taggees), and share with them information outside their norm. One woman, said, <i class="markup--em markup--p-em">“I like to tag people that I know are interested in something and whose audience will also care. But my interest is in creating a crossover. So I involve audiences that follow people in dance, but I share with them something totally different, such as poetry.” </i></div>
<div class="graf--p" name="2b6a">
<i class="markup--em markup--p-em"><br /></i>Another, told a folk story of how tags influenced Facebook’s algorithms and helped give her content more viewers, <i class="markup--em markup--p-em">“If I tag someone and then he comments, the post becomes ‘active’ and it’ll appear at the top of the News Feed. So I’ll sometimes tag someone directly in the post or in the comments. In either case, if it makes him comment, the post will be pushed to the top.”</i></div>
<div class="graf--p" name="0426">
<i class="markup--em markup--p-em"><br /></i>Recommendation algorithms have tried to filter opinions, people and items that are different to us, limiting the diffusion of information. Our study highlights how people appear to value “things outside the norm”. Our interviewees had a need to distribute fresh information that could free audiences from algorithmic biases. Our research on tagging thus raises the question — would social media benefit from more official digital structures tailored for targeting new audiences and sharing with them strange information? Audiences could be bombarded with more unwanted content, but it could also facilitate “serendipitous” discoveries. </div>
<div class="graf--p" name="f38e">
<br /></div>
<div class="graf--p" name="f38e">
To read more, checkout our <b class="markup--strong markup--p-strong">Hyptertext 2015 </b>research paper:<br />
<a class="markup--anchor markup--p-anchor" data-href="http://arxiv.org/pdf/1509.01095v1.pdf" href="http://arxiv.org/pdf/1509.01095v1.pdf"><b class="markup--strong markup--p-strong">Tag me Maybe: Perceptions of Public Targeted Sharing on Facebook</b></a>,<br />
with <a class="markup--anchor markup--p-anchor" data-href="http://www.cs.ucsb.edu/~saiph/" href="http://www.cs.ucsb.edu/~saiph/">Saiph Savage</a>, Andres Monroy-Hernandez, Kasturi Bhattacharjee, Tobias Hollerer.</div>
<div class="graf--p graf--empty" name="d992">
<br /></div>
<div class="graf--p graf--empty" name="c411">
<br /></div>
<div class="graf--p graf--empty" name="d18f">
<br /></div>
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-52402126777227002942015-03-31T20:56:00.000-06:002019-12-10T09:56:34.928-07:00#FixIT@UNAM: How We Got Over A Hundred Latinas In Our Hackathon<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="https://1.bp.blogspot.com/-zXYLjuKa8FM/Xe79VOXMzfI/AAAAAAAAgGA/3pAzVaHI1e0Fep6LqsLRUXG4R-_NuY4KQCNcBGAsYHQ/s1600/fix.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="540" data-original-width="960" height="225" src="https://1.bp.blogspot.com/-zXYLjuKa8FM/Xe79VOXMzfI/AAAAAAAAgGA/3pAzVaHI1e0Fep6LqsLRUXG4R-_NuY4KQCNcBGAsYHQ/s400/fix.jpg" width="400" /></a></div>
<br />
[Image Caption: Group photo of #FixIT: Latina Hackathon @ UNAM<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://1.bp.blogspot.com/-FPKKlZCEoWY/VRunYlkHWUI/AAAAAAAANAk/vpSPa9lvXCk/s1600/d.JPG" imageanchor="1" style="clear: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" height="266" src="https://1.bp.blogspot.com/-FPKKlZCEoWY/VRunYlkHWUI/AAAAAAAANAk/vpSPa9lvXCk/s400/d.JPG" width="400" /></a></div>
<br />
<b><span style="font-size: x-small;">[Image Caption: #FixIT was a success because we worked with the community to </span></b><br />
<b><span style="font-size: x-small;">make Latinas feel welcomed] </span></b><br />
<b><br /></b>
<b><br /></b>
<b>Highlights</b><br />
<br />
<ul>
<li>Organized <a href="https://events.withgoogle.com/fixit/">a large scale Latina Hackathon</a> #FixIT with almost 300 registered females and 147 female Hackers on February 26-27 2015 at the National Autonomous University of Mexico (UNAM).</li>
<li>The Hackathon had a large female turnout because we created a space welcoming for Latinas.</li>
<li> Hackers designed innovative smart home appliances using arduinos, and collaborated with other very diverse females, such as high school students, young professionals, graduate and undergraduates in Computer Engineering, Mechatronics, Industrial Design, Philosophy, Politial Science, among others.</li>
<li>The Hackathon's organization had powerful team work between Google Anita Borg Scholars, US and Mexican universities (UNAM, UCSB, UC Berkeley, UCD, Georgia Tech) and non-profits (Major League Hacking, SocialTIC).</li>
</ul>
<br />
<br />
Many organizations have been fighting to bring diversity into IT. But it has been a sloooow loooong progress. I don't like slow! I decided to take things into my own hands :) So I did what you always do when you’re a little guy facing a terrible future with long odds and little hope of success: I teamed up with my friends.<br />
<br />
I was very fortunate to have been awarded to join the planning committee of <b>Google Anita Borg Scholars.</b> This new organization has the mission to increment world-wide the number of women participating in computer science. I started to work closely with <a href="https://www.andrew.cmu.edu/~mkas/">Miray Kas</a>, and <a href="https://www.linkedin.com/pub/sarah-safir/13/200/abb">Sarah Safir</a>. These two women gave me the motivation and inspiration to start dreaming and making a reality: the idea of Fixing IT, rapidly bringing diversity. It gave me the drive to stay up multiple nights with friends further devising a plan. <br />
<br />
Online I got introduced to many new great people, from <a href="http://uk.linkedin.com/pub/rohan-lamba/19/633/271">other Googlers</a> who had careers focused in fostering diversity, to <a href="https://twitter.com/jpflores9">19 year old UNAM s</a>tudents, like Juan Pablo Flores, who organized <a href="http://news.mlh.io/mexican-hackathons-come-mlh-la-liga-mexicana-de-hackatones-08-25-2014">large scale Hackathons</a> throughout their country. I’ve worked with some of the strongest teams in industry, academia, and even sport (Olympic medalists!) But I’ve never seen anything like this.<br />
Starting from literally nothing, we went to having a fast moving plan: We would do a Hackathon! A female Hackathon, a Latina Hackathon. We would be fighting to incorporate two minorities at once!<br />
<br />
<b><span style="font-size: x-small;">[Image Caption: Latinas in </span></b><br />
<b><span style="font-size: x-small;">#FixIT hacking away.]
</span></b><br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://1.bp.blogspot.com/-UtcexKcj_PE/VRukuPNrp6I/AAAAAAAANAI/aR63pSoWCek/s1600/_SAM2412.jpg" imageanchor="1" style="clear: left; float: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" src="https://1.bp.blogspot.com/-UtcexKcj_PE/VRukuPNrp6I/AAAAAAAANAI/aR63pSoWCek/s320/_SAM2412.jpg" /></a></div>
<br />
<b><span style="font-size: large;">Why a Hackathon?</span></b><br />
We saw Hackathons as a fast way to turn anyone into creators, not just consumers of, technology. We wanted to create a space that would empower these minorities to build and present their own visions of what technology looks like.<br />
<br />
We were a strong motivated team, but in a certain sense we were also just newbies, running around telling everyone we were going to create this new awesome more inclusive hackathon. You can ask anyone, I was constantly inviting everyone to join and participate in my Hackathon.
The problem was that we didn't really know any latinas interested in tech; we didn't have a space to hold the Hackathon. Heck, we didn't even have a budget. We had no funds to even run it. Even I began to doubt myself. It was a rough period.<br />
<br />
But December hit and suddenly things started moving. Suddenly all the work we had done started coming together. All the folks we talked to about it suddenly began getting really involved and getting others involved. Everything started snowballing. It happened so fast.<br />
<br />
<b><span style="font-size: large;">So... How did we make it happen?</span></b><br />
<b><span style="font-size: large;"><br /></span></b>
<b><span style="font-size: large;"><br /></span></b>
<b><span style="font-size: x-small;">[Image Caption: UNAM Professors helping to explain Hackathon dynamics. <br />The professors helped to secure the spaces, </span></b><b><span style="font-size: x-small;">prepare bootcamps, and connect with students.]</span></b><br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://1.bp.blogspot.com/-MLgYQ1RZ9vw/VRuo32N9-uI/AAAAAAAANAs/uVlNLQ9IsDQ/s1600/_SAM2299.JPG" imageanchor="1" style="clear: left; float: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" height="225" src="https://1.bp.blogspot.com/-MLgYQ1RZ9vw/VRuo32N9-uI/AAAAAAAANAs/uVlNLQ9IsDQ/s1600/_SAM2299.JPG" width="400" /></a></div>
<b><span style="font-size: large;"><br /></span></b>
<br />
<div class="separator" style="clear: both; text-align: center;">
<br /></div>
<br />
<b>A. Professors from the National Autonomous University of Mexico (UNAM) </b><b>got involved.</b><br />
<span style="color: #666666;">[Note that UNAM is one of the largest public universities in the world, and also one with high academic ranking] </span>The professors played a key role in the Hackathon's success because they helped us secure a space where we could run the Hackathon. From interviews with latinas we found that: 1) some latinas did not attend Hackathons because they lacked their own equipment; 2) some parents from latinas only allowed their daughters to participate in the Hackathon if it happened during the school week and it was part of a school activity; 3) some Latinas felt insecure of their technical skills and this made them hide away from hackathons<br />
<ol>
</ol>
This lead us to want to create a Hackathon that covered the needs of these girls, and made it easy for them to participate. <br />
1. The UNAM professors helped us secure computer labs so that participants would not need to bring their own equipment.<br />
<br />
2. The professors also gave us access to their network of high school and college professors (UNAM also has high schools!) These professors kindly invited their Latina students to our event; many physically drove them there; and some even turned it into an official school activity.<br />
<br />
<b><span style="font-size: x-small;">[Image Caption: Latinas with Mentors <br />@ #FixIT bootcamp.]</span></b><br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://1.bp.blogspot.com/-yuJmZuZ-Zik/VRumba3LLzI/AAAAAAAANAc/VNXSKfjCw7I/s1600/_SAM2316.jpg" imageanchor="1" style="clear: left; float: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" height="320" src="https://1.bp.blogspot.com/-yuJmZuZ-Zik/VRumba3LLzI/AAAAAAAANAc/VNXSKfjCw7I/s1600/_SAM2316.jpg" width="180" /></a></div>
<span style="text-align: center;"><br /></span>
<span style="text-align: center;">3. The UNAM professors and also UNAM senior students helped us to prepare a bootcamp. We felt the bootcamp would give participants more security in themselves and help them thrive. The bootcamps helped participants regardless of their programming knowledge or skill to start designing and building smart houses using Arduino. UNAM also gave us many mentors that helped participants throughout the hackathon. These mentors were very engaged students that also helped to create security in participants.</span><br />
<div>
<br /></div>
<div>
From interviews, we also suddenly realized that this might be one of the first large scale latina Hackathon in the region. It then made sense to try to help participants to build a network, and inspire and help each other to succeed. We thus decided to have talks for the Hackathon. Female professors, such as <a href="http://web.cs.ucdavis.edu/~rubio/">Dr. Cindy Rubio</a> from UC Davis, and graduate students, such as <a href="http://deanab.com/">Deana Brown</a> from Georgia Tech. They all came down to UNAM to give inspiring talks on their research, career paths, and how they managed to succeed. Here again the UNAM professors managed to secure for us a large auditorium for the talks.<br />
**For all of this thanks to <a href="http://profesores.fi-b.unam.mx/normaelva/">Prof. Norma Elva Chavez </a>and <a href="http://biorobotics.fi-p.unam.mx/">Dr. Jesus Savage. </a><br />
<br />
<b style="text-align: center;"><br /></b>
<b style="text-align: center;">B. Amazing Google Anita Borg Scholars Support</b><br />
The network of Google Anita Borg Scholars, specifically Miray Kas, connected me with a group of very passionate Googlers, e.g., Rohan Lamba, who had the vision of creating large scale hackathons across the world. The group helped cover all of the costs of the hackathon, and helped immensely with the planning and logistics. Rohan also came to the hackathon at UNAM and was up with me since 4 am the day of the Hackathon planning and hacking away! Thanks to Google, our Hackathon gave participants:<br />
<br />
1) Awesome Swag (Tshirts, bags, stickers, <a href="http://www.techgirlsaresuperheroes.org/">Tech Girls are Superheroes book</a>! **This was actually written and created by another Google Anita Borg Scholar, Jenine Beekhuyzen!)<br />
2) Kits to build and design Smart Houses using arduino.<br />
3) Lots of great delicious Mexican food!<br />
4)Awesome Prizes! (This was important to motivate participants. )<br />
<br />
The Google Anita Borg Scholar network also gave us inspiring speakers and mentors. Both Dr. Cindy Rubio and Deana Brown were Anita Borg Scholars. They came down to UNAM to give inspiring strong talks.<br />
<br />
<b><span style="font-size: x-small;">[Image Caption: Dr. Cindy Rubio was the <br />keynote speaker of #FixIT]</span></b><br />
<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://1.bp.blogspot.com/-amjr28yQNK8/VRuleyj0CmI/AAAAAAAANAQ/jF77SOMI_Mk/s1600/_SAM2284.JPG" imageanchor="1" style="clear: left; float: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" height="320" src="https://1.bp.blogspot.com/-amjr28yQNK8/VRuleyj0CmI/AAAAAAAANAQ/jF77SOMI_Mk/s1600/_SAM2284.JPG" width="180" /></a></div>
<b>C. Awesome UNAM Students</b><br />
Our Hackathon had a large turnout, perhaps one of the largest Latina Hackathons to date. How did this happen? I believe it helped us immensely to listen to the needs of these young latinas. Undergraduate students from UNAM connected us with large networks of females who explained to us what it would take to have them attend the hackathon. We created spaces that facilitated their participation and they came! We had participation from schools and universities across the globe. Some examples were Universidad nacional Autonoma de Mexico (UNAM); Instituto Politecnico Nacional (IPN); ITAM; Tec de Monterrey; Georgia Tech; University of California Santa Barbara, UC Berkeley; UC Davis; Oxford.<br />
<br />
The biggest takeaway from this is that perhaps the success of our Hackathon came not necessarily because we had big tech players on our side. We succeed because there was this enormous mental shift in the students, professors, and just everyone involved. All of them were thinking of ways they could help to recruit and bring their latina friends to our event. Often really clever, ingenious ways. Students made videos. They designed ads. They bought billboards. They announced massively and repeatedly on social media. They had rallies. Students saw it as their responsibility to help.<br />
Here a big shoutout to Juan Pablo Flores, Alejandra Monroy!<br />
<br />
<b><br /></b>
<b><br /></b>
<b><br /></b>
<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://3.bp.blogspot.com/-aWnaGGhXLas/VRu5SHCIooI/AAAAAAAANBE/I7B5E25DO_w/s1600/_SAM2224.jpg" imageanchor="1" style="clear: left; float: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" height="225" src="https://3.bp.blogspot.com/-aWnaGGhXLas/VRu5SHCIooI/AAAAAAAANBE/I7B5E25DO_w/s1600/_SAM2224.jpg" width="400" /></a></div>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b><br /></b>
<b>D. The Designers</b><br />
Laura Ballroom and Ariadna Gómez Dessavre were the designers of the Hackathon. Ari made the awesome #FixIT logo; and Laura designed the tshirts, stickers and certificates.<br />
<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://2.bp.blogspot.com/-bpo5LCEWOpk/VRuqIh9NKnI/AAAAAAAANA0/RZ1g7DcPm5c/s1600/Screen%2BShot%2B2015-04-01%2Bat%2B1.18.50%2BAM.png" imageanchor="1" style="clear: right; float: right; margin-bottom: 1em; margin-left: 1em;"><img border="0" height="320" src="https://2.bp.blogspot.com/-bpo5LCEWOpk/VRuqIh9NKnI/AAAAAAAANA0/RZ1g7DcPm5c/s1600/Screen%2BShot%2B2015-04-01%2Bat%2B1.18.50%2BAM.png" width="176" /></a></div>
<br />
The designers were key for the event because they created products that were used and adopted by participants. This helped to create a collective identity for the event and build a community. I loved seeing on social media people share selfies with their Hackathon tshirts and feel proud to wear them.<br />
<br />
I loved seeing the senior female student add our sticker to her laptop (along with the multiple other stickers she proudly shows to present all the tech events she has attend) and the young high school student add our sticker to her bare laptop (representing this is her first tech events. But seems to be likely to attend now more :)<br />
<br />
<br />
I believe we won this fight, we were able to have a hackathon with hundreds of latinas; rapidly bringing diversity into technology, because everyone made themselves the hero of their own story. Everyone took it as their job to have an event where more audiences could be included; where more voices could be empowered to design and construct technology. They threw themselves into it. They did whatever they could think of to do. They didn’t stop to ask anyone for permission.<br />
<br />
We, the people decided to make a difference. We decided to make it our responsibility to do this work. To change who is empowered to participate in a Hackathon. Who is empowered to define technology. Let's not forget that we do have the power to change our reality. <br />
"Juntas Creando y Transformando la Realidad"<br />
<span style="color: #666666;">[This was the main quote for #FixIT, it reads in Spanish: Together Creating and Transforming Reality]</span><br />
<br />
<br />
<b><span style="font-size: x-small;">[Image Caption: #FixIT was a success because we worked with the community to </span></b><br />
<b><span style="font-size: x-small;">make Latinas feel welcomed] </span></b><br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://1.bp.blogspot.com/-FPKKlZCEoWY/VRunYlkHWUI/AAAAAAAANAk/vpSPa9lvXCk/s1600/d.JPG" imageanchor="1" style="clear: left; float: left; margin-bottom: 1em; margin-right: 1em;"><img border="0" height="400" src="https://1.bp.blogspot.com/-FPKKlZCEoWY/VRunYlkHWUI/AAAAAAAANAk/vpSPa9lvXCk/s1600/d.JPG" width="600" /></a></div>
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
<br />
This text is a remix of Aaron Swartz's inspiring speech: How we Stopped SOPA." We miss you Aaron. </div>
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-11682244827077922522014-11-17T10:17:00.001-07:002014-11-17T10:17:17.918-07:00Visualizing Online Audiences<div class="separator" style="clear: both; text-align: center;">
</div>
<br />
<div class="separator" style="clear: both; text-align: center;">
</div>
<br />
<div class="separator" style="clear: both; text-align: center;">
</div>
<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://4.bp.blogspot.com/-e3T__rtWB5o/UxdQkO94oEI/AAAAAAAAHhM/IHlXo3PEQvI/s1600/socialGroup.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="http://4.bp.blogspot.com/-e3T__rtWB5o/UxdQkO94oEI/AAAAAAAAHhM/IHlXo3PEQvI/s1600/socialGroup.png" height="300" width="550" /></a></div>
<b> Caption:</b> Social Spread Interface lets people select audiences based on their social connections<br />
<br />
We tend to think that the larger our online audience, the more comments and interactions our content will receive. Yet this is not always the case.<br />
<br />
The presence of other people can create<a href="http://en.wikipedia.org/wiki/Bystander_effect"> a diffusion of responsibility</a>. People don't feel as pressured to take action when they feel they share responsibility with others. Paradoxically, when someone poses a question to her entire network, her friends are less likely to respond than if she made the question to a <a href="http://research.microsoft.com/en-us/um/people/teevan/publications/papers/icwsm11-et.pdf">small targeted audience. </a>Directing content to specific groups of people can help users harvest richer online interactions. <br />
<br />
Many tech-savvy users use different sharing mechanisms to engage in<i> <a href="http://research.google.com/pubs/archive/37843.pdf">selective sharing</a>, </i>directing content to specific predefined audiences. These users usually define list of people with particular interests or social ties (e.g., coworkers.) They then post content contextualized so that it is relevant to the interests of the people in each list.
<br />
But, keeping up-to-date lists can be hard and time-consuming. It's also inapplicable for more dynamic interactions, based on location, or popularity of the targeted users. For example, someone organizing a social rally might only want to target the friends who are in town on a particular day. Or a person who just wrote an article on Gay rights might want the help from their most influential friends in the topic to promote their article.<br />
Another technique involves selecting individuals to target on-the-fly and only sharing the content or message to them.
This type of behavior allows for a more dynamic selective sharing experience that is context-driven. This practice is usually referred to as <i>targeted sharing</i>.<br />
<br />
Finding the right people at the right time is hard, especially when we are interacting in large communities where it is hard to keep track of everyone's interests and traits. Previous work, including <a href="https://www.facebook.com/about/graphsearch">Facebook's graph search</a>,
used list-based interfaces to recommend people with a certain expertise, interest or trait. But these systems do not let people easily explore and compare the different characteristics of the recommended individuals. However, these characteristics <a href="https://www.blogger.com/blog.leif.me/2012/06/mutual-assessment/%E2%80%8E">can play an important role</a> when people are deciding whether or not to select a person for a particular collaboration or interaction. People want ways through which they can understand the space of users they can target to interact. List based interfaces don't let people have quick overviews of their possible targeted audience. Nor do they let people easily zoom in and compare specific users.
But to have rich sharing experiences people need to understand their audience: an overview and its details. For instance, a person wanting to post on LGBT might need to understand that half of her interested audience is from Russia. This might mean that to better engage with them, she should include some LGBT issues happening in their country.<br />
<br />
Interactive visualization tools can enable effective audience targeting by prompting a user to learn about their audience and to understand their different interests. To explore these ideas, we designed <i>Hax</i>. Hax is a tool that provides a query interface and multiple visualizations to support users in
dynamically choosing audiences for their targeted sharing tasks. We study how users engaged with this tool in the context of sharing and connecting with an audience on Facebook.<br />
<br />
We believe the data modeling techniques that work for content categorization and information retrieval can be adapted to mine people's interests and retrieve audiences relevant to users' diverse needs. But, while specialized data modeling algorithms exist that can correctly categorize data, they rarely fully capture the complex and ever-changing decision-making process for targeting an audience. We therefore opt to integrate data visualizations that incorporate a human-in-the-loop approach.
We designed different data visualizations that highlight specific traits, or social signals, of relevant individuals in order to aid users in their audience targeting tasks.<br />
<br />
Our exploration begins with the three social signals listed below. We briefly define the signal and the reasons for considering it. We decided to begin with these signals as previous work identified they play an important role in targeting audiences:<br />
<br />
<b>Shared interests:</b>
This signal captures the personal thematic interests of each community member.
<a href="http://alejandrogg.com.mx/temario3/Ann-collaboration%20process.pdf%E2%80%8E">Many researchers</a> and practitioners view collaborations as a process that aggregates personal interests
into collective choices through self-interested bargaining. We believe this bargaining process can be
facilitated by making users aware of the personal interests of others, and how they relate to the collaboration task they are promoting.<br />
<br />
<b>Location:</b> This signal holds information about the countries, states, and cities where community
members live. Collaborations supported by computers have traditionally provided users with the
luxury of interacting with others without having to worry about their location. However, location does play an <a href="http://www.grantmckenzie.com/academics/McKenzieG_COSIT2011.pdf%E2%80%8E">important role </a>when interacting and organizing events within the physical
world (e.g., a social rally) as others' spatial-temporal constraints can determine <a href="http://www.cs.ucsb.edu/~holl/pubs/Savage-2011-LBS.pdf">how mucha person will engage in the activity</a>.<br />
<br />
<b>Social Spread:</b> This signal holds information about the type of friends and social ties community members have. This signal is important because it can aid members in recognizing
prospective newcomers who can <a href="http://repository.cmu.edu/cgi/viewcontent.cgi?article=1089&context=hcii%E2%80%8E">help keep the community alive and active.</a> Additionally, the social connections of a member can also help in the spread of the community's messages and visions.
Members could thus use this signal to identify the users whose social connectivity would help them the most in distributing certain content.<br />
<div class="separator" style="clear: both; text-align: center;">
</div>
<div class="separator" style="clear: both; text-align: center;">
<a href="http://3.bp.blogspot.com/-QswpTk7q-dk/UxdLxpVK3jI/AAAAAAAAHhA/XC5B5gP3ZnU/s1600/interfaces.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="http://3.bp.blogspot.com/-QswpTk7q-dk/UxdLxpVK3jI/AAAAAAAAHhA/XC5B5gP3ZnU/s1600/interfaces.png" height="640" width="364" /></a></div>
<div class="separator" style="clear: both; text-align: center;">
<br /></div>
<span style="font-size: x-small;"> </span><span style="text-align: center;"><b>Caption: </b>Zoomed in Version of our Location Based Interface (top,) Transparent Interface (middle,) and Social Spread (bottom)</span><br />
<div>
<span style="text-align: center;"><br /></span></div>
<div class="separator" style="clear: both; text-align: center;">
<span style="color: #0b5394;"><i style="background-color: white; font-family: Georgia, 'Times New Roman', Times, serif; font-size: 13px; line-height: 22px; text-align: left;">For more, see our full paper to be presented at COOP2014, <b style="text-decoration: none;"><a href="https://www.dropbox.com/s/d6zifb1umo6846o/visualizingTargetedAudiences.pdf">Visualizing Targeted Audiences</a> </b>co-authored by </i><i style="background-color: white; font-family: Georgia, 'Times New Roman', Times, serif; font-size: 13px; line-height: 22px; text-align: left;"><a href="http://www.cs.ucsb.edu/~saiph/"><b>S</b>aiph Savage</a>, <a href="http://angusforbes.com/">Angus Forbes</a>, <a href="http://scholar.google.com/citations?user=u1PqC9QAAAAJ&hl=en">Carlos Toxtli,</a> <a href="http://www.grantmckenzie.com/academics.html">Grant McKenzie,</a> and <a href="http://ccl.angusforbes.com/index.php?people">Shloka Desai</a></i><i style="background-color: white; font-family: Georgia, 'Times New Roman', Times, serif; font-size: 13px; line-height: 22px; text-align: left;"></i></span></div>
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-22129099194309369172013-10-14T00:51:00.004-06:002014-12-22T19:11:43.484-07:00You know that thing I used to hate?...I love it now!: Expert Evolution in Social MediaIt took me a long time to really appreciate some of the most successful people that have ever lived, such as Steve Jobs or Coco Chanel. Despite the adversity of being born orphans, these individuals changed their reality and transformed the world that we know.<br />
<br />
Have you ever felt that one same thing, can have various different meanings in different points of your life? For example, I used to dislike Apple products, due to their closed architecture, lack of customization, all the related Steve Jobs fanboys etc. Now my opinion has shifted, and I am considering purchasing a macbook pro.<br />
<br />
This paper: <b><a href="http://cs.stanford.edu/people/jure/pubs/beerrec-www13.pdf">"From Amateurs to Connoisseurs: Modeling the Evolution of User Expertise through Online Reviews"</a> </b>from www2013, models precisely how users' perceptions on products (particularly beer!) change through time. Note that perceptions can be compared to level of expertise. <br />
The work considers that users evolve on their own "personal clock" (i.e., each user is modelled on its own time-scale.) so some users may be really experienced when they provide their first review, while other users may never ever become experts. It is also assumed, that users with similar levels of expertise
will rate products in similar ways, even if their ratings are temporally far apart, e.g., a user from 2013 will rate things similarly to a user from 1999, if both users are in the same "noob level" (note that what will remain constant throughout time is what the "noob level" means.)<br />
<br />
The work models the level of experience of each user as a
series of latent parameters that are constrained to be monotonically
non-decreasing as a function of time. Therefore users can only become more experienced (or stay at least at the same level of experience...this clearly does not consider cases related to the German word of "Verlernen," where you forget something you have already learned!) Users evolve as they rate more products, i.e. the work considers that the very act of consuming products will cause users' tastes to change.<br />
In particular the authors consider that users move from one level of expertise to another, depending on how they rate particular products. Given that the paper works with a beer rating dataset, the authors consider that expert users will rate higher the hoppiest beers, i.e. the strong Ales (confused what hops in beer mean, <a href="http://en.wikipedia.org/wiki/Hops">check this article out.</a>) It is assumed that liking strong Ales is an acquired taste. <br />
The figure below shows the ratings different users have given to different beers. We see how the Firestone XV, a strong ale, is one of the beer that was rated the highest, and they consider this corresponds to an expert rating. The figure also shows how biases can exist for different beers given the level of expertise of the user. <br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://2.bp.blogspot.com/-lf2Xh41uibc/UlwZOcD9c1I/AAAAAAAAEbc/4flrgpnHOwQ/s1600/www.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="http://2.bp.blogspot.com/-lf2Xh41uibc/UlwZOcD9c1I/AAAAAAAAEbc/4flrgpnHOwQ/s320/www.png" height="289" width="320" /></a></div>
<br />
The approach is somewhat simple, different recommender systems are created for different stages of user evolution.<br />
A classic latent factor recommender system would look something like the following: <br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://2.bp.blogspot.com/-vd67d6FtKSI/UlwjFAwvMmI/AAAAAAAAEb0/UvUsZIp1hOA/s1600/www2.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="http://2.bp.blogspot.com/-vd67d6FtKSI/UlwjFAwvMmI/AAAAAAAAEb0/UvUsZIp1hOA/s320/www2.png" height="101" width="320" /></a></div>
<div class="separator" style="clear: both; text-align: center;">
</div>
<br />
The authors create a sort of feature vector, that has different recommender systems for each stage of the user's life cycle:<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://3.bp.blogspot.com/-j9STbjKMmss/UlwkGxGVPnI/AAAAAAAAEcA/qA4H-gxl_a8/s1600/www3.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="http://3.bp.blogspot.com/-j9STbjKMmss/UlwkGxGVPnI/AAAAAAAAEcA/qA4H-gxl_a8/s320/www3.png" height="67" width="320" /></a></div>
<br />
<br />
Latent factor recommender models, are a collaborative filtering type recommendation algorithms, which consider that for specific domains (e.g., action films or beer) there exits a set of factors that influence the rating a specific item receives. These factors are not obvious, and it is difficult to predict the actual impact they have on a rating.<br />
The first goal of the latent factor recommender models is to infer these latent factors from the data by using mathematical techniques. <br />
<br />
In the author's approach it is considered that users that have a certain experience level will be influenced by certain type of factors in their rating. So e.g., a novice user the fact that a certain beer is more inexpensive than another might play a big role in how much the user says to like the beer, but for a more experiencied user, he might be more influenced by the beer's texture. This is why the authors consider different recommender models for each experience level of the user.<br />
<br />
Latent Factor models map users and items into a latent feature space.<br />
A user's feature vector denotes the user's affinity to each of the features. The product or item's feature vector represents how much the item itself is related to the features. A rating is approximated by the dot product of the user feature vector and the item feature vector.<br />
Specifically, consider that we have a set <img alt="U" class="latex" src="http://www.quuxlabs.com/wp-content/latex/4c6/4c614360da93c0a041b22e537de151eb-ffffff-000000-0.png" title="U" /> of users, and a set <img alt="D" class="latex" src="http://www.quuxlabs.com/wp-content/latex/f62/f623e75af30e62bbd73d6df5b50bb7b5-ffffff-000000-0.png" title="D" /> of items. Let <img alt="\mathbf{R}" class="latex" src="http://www.quuxlabs.com/wp-content/latex/e1f/e1fd601dbae82a538d518550acb1af19-ffffff-000000-0.png" title="\mathbf{R}" /> of size <img alt="|U| \times |D|" class="latex" src="http://www.quuxlabs.com/wp-content/latex/c66/c664f3d3a2c99618e6cd3586e9bd3ce9-ffffff-000000-0.png" title="|U| \times |D|" />
be the matrix that contains all the ratings that the users have given to the items.<br />
The first task of this method is to discover the $K$ latent features. At the end of the day, we want to find two matrics matrices U (a <img alt="|U| \times K" class="latex" src="http://www.quuxlabs.com/wp-content/latex/b61/b6124fad67425b33363bd7ef9d4fe920-ffffff-000000-0.png" title="|U| \times K" /> matrix) and M (a <img alt="|D| \times K" class="latex" src="http://www.quuxlabs.com/wp-content/latex/ebb/ebb59df8dc321a9ddfc458db87dfb839-ffffff-000000-0.png" title="|D| \times K" /> matrix) such that their product approximates <img alt="\mathbf{R}" class="latex" src="http://www.quuxlabs.com/wp-content/latex/e1f/e1fd601dbae82a538d518550acb1af19-ffffff-000000-0.png" title="\mathbf{R}" />: <br />
<br />
The following image shows how the rating of items is composed of the dot product of matrices with these feature vectors:<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://2.bp.blogspot.com/-5nAT2qgy3KQ/UmVgwmK07HI/AAAAAAAAEjE/KlLemrDdjpg/s1600/www2014.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="http://2.bp.blogspot.com/-5nAT2qgy3KQ/UmVgwmK07HI/AAAAAAAAEjE/KlLemrDdjpg/s320/www2014.png" height="80" width="320" /></a></div>
<br />
Note that the M transposed matrix corresponds to the feature vectors of items.<br />
Each row of U represents the strength of the associations between a user and the features. Similarly, each row of M represents the strength of the associations between an item and
the features. To get the prediction of a rating of an item <img alt="d_j" class="latex" src="http://www.quuxlabs.com/wp-content/latex/fb1/fb1793a0a1f0a7f569eaaceb6bd6e7ff-ffffff-000000-0.png" title="d_j" /> by <img alt="u_i" class="latex" src="http://www.quuxlabs.com/wp-content/latex/eb0/eb00a04135562ae6f74786f084f54327-ffffff-000000-0.png" title="u_i" />, we calculate the dot product of the two vectors corresponding to <img alt="u_i" class="latex" src="http://www.quuxlabs.com/wp-content/latex/eb0/eb00a04135562ae6f74786f084f54327-ffffff-000000-0.png" title="u_i" /> and <img alt="d_j" class="latex" src="http://www.quuxlabs.com/wp-content/latex/fb1/fb1793a0a1f0a7f569eaaceb6bd6e7ff-ffffff-000000-0.png" title="d_j" />:<br />
<br />
<center>
<br />
<img alt="\hat{r}_{ij} = p_i^T q_j = \sum_{k=1}^k{p_{ik}q_{kj}}" class="latex" src="http://www.quuxlabs.com/wp-content/latex/864/864245053b03c6e746cbe83922746032-ffffff-000000-1.png" title="\hat{r}_{ij} = p_i^T q_j = \sum_{k=1}^k{p_{ik}q_{kj}}" /><br />
</center>
<br />
<br />
<br />
<br />
The task is now to obtain U and M.
One common way to approach this problem is the first intialize the two
matrices with some values, calculate how `different’ their product is to R,
and then try to minimize this difference iteratively. This method is called gradient descent. Its purpose is find a local minimum of the
difference. <br />
This difference is usually called the error between the estimated
rating and the real rating, can be calculated by the following equation
for each user-item pair: <br />
<center>
<br />
<img alt="e_{ij}^2 = (r_{ij} - \hat{r}_{ij})^2 = (r_{ij} - \sum_{k=1}^K{p_{ik}q_{kj}})^2" class="latex" src="http://www.quuxlabs.com/wp-content/latex/09b/09b9fe29e79709943fbe8abc12d41025-ffffff-000000-1.png" title="e_{ij}^2 = (r_{ij} - \hat{r}_{ij})^2 = (r_{ij} - \sum_{k=1}^K{p_{ik}q_{kj}})^2" /><br />
</center>
<br />
The squared error is considered because the estimated rating can be either higher or lower than the real rating. <br />
To minimize the error, we need to know in which direction we have to modify the values of <img alt="p_{ik}" class="latex" src="http://www.quuxlabs.com/wp-content/latex/dec/dec680a5862b2125d6c2757ace17beba-ffffff-000000-0.png" title="p_{ik}" /> and <img alt="q_{kj}" class="latex" src="http://www.quuxlabs.com/wp-content/latex/ed9/ed970bb4b6c725588b409188d7170f92-ffffff-000000-0.png" title="q_{kj}" />.
In other words, we need to know the gradient at the current values. Thus we differentiate the above equation with respect to these two
variables separately:<br />
<br />
<center>
<br />
<img alt="\frac{\partial}{\partial p_{ik}}e_{ij}^2 = -2(r_{ij} - \hat{r}_{ij})(q_{kj}) = -2 e_{ij} q_{kj}" class="latex" src="http://www.quuxlabs.com/wp-content/latex/ef6/ef6ddecfa568a7882395b03d9a43b98e-ffffff-000000-1.png" title="\frac{\partial}{\partial p_{ik}}e_{ij}^2 = -2(r_{ij} - \hat{r}_{ij})(q_{kj}) = -2 e_{ij} q_{kj}" /><br />
<img alt=" \frac{\partial}{\partial q_{ik}}e_{ij}^2 = -2(r_{ij} - \hat{r}_{ij})(p_{ik}) = -2 e_{ij} p_{ik}" class="latex" src="http://www.quuxlabs.com/wp-content/latex/dcf/dcf543615b34c3c92c52ac570697ddb1-ffffff-000000-1.png" title=" \frac{\partial}{\partial q_{ik}}e_{ij}^2 = -2(r_{ij} - \hat{r}_{ij})(p_{ik}) = -2 e_{ij} p_{ik}" /><br />
</center>
<br />
With the gradient, the update rules for both <img alt="p_{ik}" class="latex" src="http://www.quuxlabs.com/wp-content/latex/dec/dec680a5862b2125d6c2757ace17beba-ffffff-000000-0.png" title="p_{ik}" /> and <img alt="q_{kj}" class="latex" src="http://www.quuxlabs.com/wp-content/latex/ed9/ed970bb4b6c725588b409188d7170f92-ffffff-000000-0.png" title="q_{kj}" /> can be formulated as:<br />
<br />
<center>
<br />
<img alt="p'_{ik} = p_{ik} + \alpha \frac{\partial}{\partial p_{ik}}e_{ij}^2 = p_{ik} + 2\alpha e_{ij} q_{kj} " class="latex" src="http://www.quuxlabs.com/wp-content/latex/d8a/d8a4070c59529c9e13036a6bafaabdf8-ffffff-000000-1.png" title="p'_{ik} = p_{ik} + \alpha \frac{\partial}{\partial p_{ik}}e_{ij}^2 = p_{ik} + 2\alpha e_{ij} q_{kj} " /><br />
<img alt="q'_{kj} = q_{kj} + \alpha \frac{\partial}{\partial q_{kj}}e_{ij}^2 = q_{kj} + 2\alpha e_{ij} p_{ik} " class="latex" src="http://www.quuxlabs.com/wp-content/latex/c8c/c8c559c69dab539af106d84538f27a4f-ffffff-000000-1.png" title="q'_{kj} = q_{kj} + \alpha \frac{\partial}{\partial q_{kj}}e_{ij}^2 = q_{kj} + 2\alpha e_{ij} p_{ik} " /><br />
</center>
<br />
<img alt="\alpha" class="latex" src="http://www.quuxlabs.com/wp-content/latex/7b7/7b7f9dbfea05c83784f8b85149852f08-ffffff-000000-0.png" title="\alpha" /> is a constant that helps determine the rate of approaching the minimum. Usually the value of <img alt="\alpha" class="latex" src="http://www.quuxlabs.com/wp-content/latex/7b7/7b7f9dbfea05c83784f8b85149852f08-ffffff-000000-0.png" title="\alpha" /> is small to avoid the risk of missing the minimum and oscillating around the minimum. <br />
<br />
Ok...<br />
Now that latent factor models are clear (hopefully)...let's get back to the author's problem.<br />
As we mentioned previously, the authors consider that for each evolution stage of a user, there will be a recommender system that can adequtly capture the user's taste at that point in his life.<br />
So their problem comes down to:<br />
-Fitting the parameters of their recommender system.<br />
-Fit the user's experience progression (i.e., be able to state that when the author reviewed X item, they were at a certain experience rate.)<br />
<br />
Note that once we know the experience level in which all users contributed each of their reviews, fitting the parameters of each recommender model is a snitch (it's basically the same procedure experienced for classic latent factor recommendations.)<br />
<br />
The question now is, how do we fit the user's experience progression? Well if we assume that users gain experience monotonically, as they rate more products, we can fit experience using dynamic programming.<br />
<br />
<div class="separator" style="clear: both; text-align: center;">
<a href="http://1.bp.blogspot.com/-B7xkfoZxc4E/UmWAPR51KBI/AAAAAAAAEkk/Oih8ngHxU-U/s1600/w2.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="http://1.bp.blogspot.com/-B7xkfoZxc4E/UmWAPR51KBI/AAAAAAAAEkk/Oih8ngHxU-U/s320/w2.png" height="174" width="320" /></a></div>
<br />
The two steps (fitting each user's experience level, and fitting the parameters of each recommender ) are repeated until conversion. <br />
<br />
<br />
<br />Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-24536596502677375752013-06-16T00:42:00.000-06:002013-06-16T23:52:42.890-06:00Studying the Thesis of PhD Heroes: Munmun De Choudhury<div class="separator" style="clear: both; text-align: center;">
<a href="http://epictu.be/wp-content/uploads/2013/06/heath-ledger-talks-about-life.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" height="620" src="http://epictu.be/wp-content/uploads/2013/06/heath-ledger-talks-about-life.jpg" width="499" /></a></div>
Given that I am in the process of beginning to write my PhD thesis, I am currently reviewing the PhD thesis of doctors that have been more than successful in their career; these are people I admire and find inspirational for my own PhD path: my PhD heroes.<br />
I have decided to create blog posts that describe some of the main contributions that these PhD thesis had, the new ways of thinking that these doctors brought in.
I will begin this series with the PhD dissertation of <a href="http://research.microsoft.com/en-us/um/people/munmund/">Munmun De Choudhury</a>, currently working in Microsoft research; she has several publications in top conferences such as CSCW, CHI, ICWSM, WWW, among others. Munmun is indeed one of my main true PhD heroes.<br />
<br />
Her thesis focused on designing frameworks and computational models to obtain a detailed understanding of how communication happens in online social networks. It was considered that online communication patterns are divided in two main forms: the actual message discussed, and the channel or media used to discuss the message. Work before Dr. De Choudhury's thesis focused more on studying the network structure and dynamics, and little emphasis was given to providing tools that could characterize the type of messages present in an online community, providing insightful observational studies on large-scale social communication datasets.<br />
<br />
In particular, her research explored 3 main areas: (1) how information is diffused in an online social network, analyzing in particular how the influence of users and the fact that you can have very many similar users talking to each other, affects information spread; (2) how communication dynamics in online communities can be modeled, particularly focused on external and internal communication factors; (3) how "interestingness" of conversations can be modeled and measured , in particular focusing on detecting interesting conversations and identifying the features that turn them into interesting content.<br />
<br />
Providing means to explore and analyze what are the dynamics and impact of our online social communications is important because social media data has shown to originate and create real world revolutions, think e.g., elections in Iran, Earthquake in Haiti. Social media also enables viral marketing, enabling collaborations in corporations, and can help users find experts, or even people that can help them connect with others.<br />
<br />
In the following we begin exploring in detail each of the main themes discussed in her thesis.<br />
<br />
<b>Measuring the Intrestingness of a Conversation</b>: The work considered that a conversation was interesting, when it made participants return to the conversation and continue commenting and posting. Such behavior is observed frequently on youtube, when users have already watched the video, yet they are returning to the video to comment and respond to others.<br />
The work considered that people will participate and return to conversation when the theme of the conversation is engaging and/or interesting people are participating in the discussion. They predicted that users will return to a conversation, when they: (a) find the whole conversation theme interesting; (b) see comments by people that are well known in the community; (c) observe an engaging dialogue between two or more people (an absorbing back and forth between two people).<br />
Additionally conversations that are interesting will be propagated throughout the network; we will observe things like: users will seek other users who participated in interesting conversations; interesting themes will tend to be present in other conversations in the community; users who participated in the interesting conversations will search for other similar conversations about the same theme.<br />
Themes are defined as a sets of salient topics associated with conversations at different points in time.<br />
<br />
Interesting users are defined as users who after they comment, they receive a wide variety of comments from others; users that tend to participate in conversations that are currently popular in the community; users that tend to comment and engage in conversations with other interesting users.<br />
<br />
<i>Theme modeling: </i>Within the modeling of themes, an idea that I found interesting from this thesis is that while there was a focus on modeling what themes were present in a conversation in a given time period, there was also an emphasis on normalizing the amount of content that was associated with a theme based on time and based on co-participation. This helped identified themes that were not only temporally popular or interesting due to an external event, or themes that certain users tended to frequently comment, not so much because the conversations around the theme were interesting, but rather because they had a probable passion for the subject.<br />
<br />
<b>Information Difusion:</b><br />
(post in progress...come back soon!:)Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com3Mexico City, Federal District, Mexico19.4326077 -99.13320799999996818.9531097 -99.778654999999972 19.912105699999998 -98.487760999999963tag:blogger.com,1999:blog-30573458.post-27614953763940392782012-11-14T12:28:00.001-07:002014-12-22T19:12:03.055-07:00Another layman's explanation of: Expert Evolution in Online Social Networks <div class="separator" style="clear: both; text-align: center;">
<a href="http://vator.tv/images/attachments/300709083739sms229.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="http://vator.tv/images/attachments/300709083739sms229.jpg" height="400" width="500" /></a></div>
I was recently reading a very interesting paper titled: <b>Evolution of Experts in Question Answering Communities</b> by Aditya Pal, Shuo Chang and Joseph Konstan. And thought I would share the paper and intend to explain it in Layman's terms. <br />
There has been vast amount of work done in detecting experts in Question Answering Communities, typically this analysis is either through graph based methods or feature based methods. Graph based methods tend to analyze the link structure of a user in an online social network to find authoritative users. They analyze things such as: to how many other people is the user "friends" to? Feature based methods, on the other hand, analyze the characteristics of the users: how many best answers does the user have? What language style does he use? etc etc<br />
The work we are analyzing seeks to identify experts, but then does a temporal analysis, to study how experts evolve in a community and how they influence a community's dynamics. The online community studied is Stackoverflow. To identify experts, the authors used two approaches: On one hand, they identify the number of positive votes a user's answerers and questions have received (a user gets a positive vote, when his/her answer is helpful to the community, or when his/her question is interesting or relevant to someone in the community) and labeled the top 10% of users with the highest number of votes as experts.<br />
To analyze how experts evolve and how a community can be influenced in time by the answers and social interactions of experts, the authors performed the following:<br />
<ol>
<li>the questions and answers of the community were divided into bi-weekly buckets. Were the first bucket would hold the questions and anwsers of the first two weeks of the stackoverflow data they had collected, the second bucket the questions and answers created in the 3-4th weeks etc etc </li>
<li>For each user it is then possible to calculate per bucket (per every 2 weeks,) the number of questions, answers and best answers he/she have given. </li>
<li>For each user a relative time series is computed of each data type he/she has generated (questions, answers and best answers). This relative time series is constructed so that the contribution of a user can be valued relatively to the contribution of other users. For this, what is done, is that in each of the time buckets the mean and standard deviation for each data type are calculated. (lets recall that a bucket holds the number of answers, questions and best answers different users have given in that particular time period, so for each type of variables, we can calculate the mean and standard deviation. It is then possible to normalize a data point in the time bucket as: <br />X_b=(X_b - Mean_b)/(standardDeviation_b)<br /><br />Where X_b represents the number of answers a particular user has generated in time bucket b. And Mean_b represents the mean of all the number of answers different users have given in time bucket b</li>
<li>After this step, each user is associated with 3 relative time series: the time series of their answers, questions and best answers. From the answers and best answer time series, a point wise ratio between best answers and answers is then calculated. This point wise ratio indicates the probability of a user's answers being selected as the best answer. <br />The following figure shows an interesting plot where we see how the likelihood of an expert and an average user receiving the votes for best answer changes over time. </li>
</ol>
<div class="separator" style="clear: both; text-align: center;">
<a href="http://1.bp.blogspot.com/-0zFnwV8SBck/UKP_gwBbgRI/AAAAAAAACVM/HzNg2hkGrV8/s1600/timeBucket.png" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" src="http://1.bp.blogspot.com/-0zFnwV8SBck/UKP_gwBbgRI/AAAAAAAACVM/HzNg2hkGrV8/s320/timeBucket.png" height="233" width="320" /> </a></div>
<div class="separator" style="clear: both; text-align: center;">
<br /></div>
What we notice is that the likelihood of receiving the best answer increases significantly over time for experts in comparison to average users. Initially the likelihood of receiving a best answer is the same for both experts and average users. The authors believe that this occurs, because when a new person, who happens to be an expert, joins the community, other users are wary of marking the answers of newcomers as the best. But as the expert gains reputation, the rest of the community members become more and more comfortable in marking their answers as the best.
<br />
The next interesting thing the author's analyzed was the the likelihood of having a user ask a question. It was seen that in general expert users do not ask questions. They found that the overall question to answer ratio among experts was 1/15 !!! To compare the time series of questions and answers, the authors computed an aggregate time series of the number of questions and answers of experts, and then normalize the time series such that it has mean=0 and standard deviation =1. From these two resulting distributions (questions and answers) a cross-covariance was computed. Now, the cross-covariance will give us information about just how similar two signals are, as a function of a time-lag applied to them. The authors found that the optimal time lag was zero for the majority of expert users. Which indicates that likelihood of an expert asking or responding to a question vary simultaneously. Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com3tag:blogger.com,1999:blog-30573458.post-16749133385854061652012-07-27T10:24:00.001-06:002012-07-27T20:13:03.360-06:00Layman's Explanation of Online LDA<img src="https://dhs.stanford.edu/wp-content/uploads/2011/02/Four-Topic-Modules-1024x838.png" alt="Topic Modeling!" height="400" width="500" />
<br>LDA stands for Latent Dirichlet Allocation, and it is a type of topic modeling algorithm. The purpose of LDA is to learn the representation of a fixed number of topics, and given this number of topics learn the topic distribution that each document in a collection of documents has. For example, if we were given the following sentences:<br><b>
A:I spend the day at the beach tanning.<br>
B: I ate Mexican Tacos and Guacamole.<br>
C:I love tanning in Mexican beaches while eating quesadillas and tacos under the sun.</b><br><br>
LDA might say something like:<br><b>
Sentence A is 100% about Topic 1<br>
Sentence B is 100% Topic 2<br>
Sentence C is 30% Topic 1, 70% Topic 2</b> <br><br>
where LDA also discovers that: <br><b>
Topic 1: 30% beach, 15% tanning, 10% sun, … (where we notice that topic 1 represents things related to the beach)<br>
Topic 2: 40% Mexican, 10% Tacos, 10% Guacamole, 10% Quesadilla , … (where we notice that topic 2 represents things related to Mexico.)</b><br><br>
LDA learns how topics and documents are represented in the following form:<br><br>
1)First the number of topics to discover is selected. (Similar to when we specify the number of clusters we wish our clustering algorithm to consider)<br><br>
2) Once the number of topics is selected, LDA will go through each of the words in each of the documents, and it will randomly assign the word to one of the K topics. After this step we will have topic representations (how the words are distributed in each topic) and documents represented in terms of topics (Just like the above example, where we said Sentence or Document C is 30% about Topic 1 and 70% about Topic 2.) Now, the thing is, the assignment of words to topics, was done in a random form, so of course this obtained representation is not very optimal or accurate. To better this representation LDA will analyze per document: <br>
what is the percentage of words within the document that were assigned to a particular topic. And for each word in the document, LDA will analyze over all the documents, what is the percentage of times that particular word has been assigned to a particular topic. LDA will therefore be calculating:<br><br>
1) p(topic t | document d) = percentage of words in a document d that are currently assigned to topic t.<br>
2) p(word w | topic t) = percentage of times the word w was assigned to topic t over all documents.<br><br>
LDA will decide to move a word w from topic A to topic B when:<br>
p(topic A | document d) * p(word w | topic A)< p(topic B | document d) *
p(word w |topic B)<br>
After a while, LDA "converges" to a more optimal state, where topic representations and documents represented in terms of these topics are ok.<br><br>
Now that we have understood the underlining principle about how LDA works. We will now discuss online LDA.<br>
The problem with LDA is that the posterior probability we need to calculate in order to reassign words to topics is very difficult to compute. Therefore researchers use approximation techniques to find what this posterior probability is. <br> Generally algorithms for approximating this posterior probability are either based on sampling approaches or optimization approaches. Sampling approaches are typically based on Markov Chain Monte Carlo (MCMC) sampling. MCMC intends to find the posterior probability distribution by randomly drawing values from a complex distribution of interest. MCMC are named that way, because the previous sampled values (previous states) affect the generation of the next random sample value ( in other words, the transition probabilities between sample values is a function of the most recent sample value.) <br>
Optimization approaches on the other hand, are typically based on variational inference. Variational Inference can be seen as deterministic alternative to MCMC. Variational Inference replaces MCMC's random, somewhat independent sampling, with optimization. Variational Inference seeks to optimize a simplified parametric distribution to be close in Kullback-Leibler divergence to the posterior. The following picture, intends to show how variational inference defines a subfamily of distributions, and the goal is to find a point in the subfamily distribution that is the closest to P(z|x). Similarity is measured using Kullback–Leibler divergence, which is a non-symmetric measure of the difference between two probability distributions P and Q.
<div class="separator" style="clear: both; text-align: center;">
<a href="http://1.bp.blogspot.com/-fts1itG656w/UBM9oI1XayI/AAAAAAAACUY/IYCwIpDaae0/s1600/vari.png" imageanchor="1" style="margin-left:1em; margin-right:1em"><img border="0" height="242" width="400" src="http://1.bp.blogspot.com/-fts1itG656w/UBM9oI1XayI/AAAAAAAACUY/IYCwIpDaae0/s400/vari.png" /></a></div>
Variational Inference has shown to be as accurate as MCMC, but FASTER, so this has made Variational Inference very popular when applying it to large datasets. <br><br>
Now, despite the benefits Variational Inference brings. Large scale data analysis can still be difficult. What many groups have done is to use batch variational inference, where there is a constant iteration between analyzing each observation and updating dataset-wide variational parameters, but in really big datasets each iteration can become very costly and impractical...and this is where Online LDA comes to the rescue!<br>
Online LDA is based on online stochastic optimization, which has shown to produce good parameter estimates dramatically faster than batch algorithms on large datasets.<br>
Online stochastic optimization in LDA is about finding a balance between exploiting the knowledge gained on a particular topic assignation, and exploring new topic assignations.
Note: Images from standford university and princeton universityLittle Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com7tag:blogger.com,1999:blog-30573458.post-68491107470005510182012-04-01T10:04:00.010-06:002012-04-01T18:55:23.729-06:00Social Signals and Machine interpretation<img src="http://decoder.drugfree.org/wp-content/uploads/2008/07/girl-on-computer.jpg" alt="some_text" /><br /><br />Have you found yourself alone in the dark, working on your computer feeling a tad depressed and wishing your computer could respond in some way to your mental and emotional state? You know, maybe your computer could send you some funny comics to lighten your day, or perhaps send you an inspiring quote to keep you in the fight for life....These ideas might sound a bit far fetched given our current reality with our machines. But there is actually an active community, whose goal is precisely for machines to understand human emotion and social interaction.<br />Having machines understand emotion and social interactions, is beneficial for:<br /><br /><ul><li>Social Scientists, Psychologists and Doctors: As they all are very interested in observing and quantifying human behaviour. For Psychologists and Doctors this could help them in diagnosing and rehabilitating their patients.</li></ul><ul><li>Intelligent Algorithms: that by understanding their user better, could respond to semantic queries and retrieve more relevant material. As humans we are very familiar with interacting with different meanings given different contexts, think for example of that popular 60's song titled: "it's the same old song, but with a different meaning since you've been gone <!--3". A tad korny, but true, meanings change with context. Having machines that can understand this is important. </li--></li><li>Ambient intelligence: Environments can become more responsive to the social context. Perhaps the room detects that the crowd at a small reunion is bored, and so the room could maybe start playing whimsical animal figures on the walls to entertain the audience.</li></ul><br />Overall having machines being able to interpret human emotion can improve Human Computer Interaction, as it increases the computer's sensitivity to the user's mental and emotional state [5].<br /><br />Now, given that we understand the benefits of machines that can understand our emotions betters. The question is, so... how can this be enabled?<br /><br />Recent investigations focus on something called SOCIAL SIGNALS.<br /><br />But what is exactly a social signal?<br />-A <span style="font-weight:bold;">social signal</span> (According to Poggi and D'Errico) is a communicative or informative signal that conveys information about social actions, social interactions, social emotions and attitudes.<br /><br />Where the heck does this idea of "Social Signal" come from?<br />The term "Social Signal" was inspired by various psychology studies, that analysed how non-verbal behaviour relates to social interactions. Psychologists were studying things such as:<br /><br /><ul><li>How does non-verbal behaviour effect the formation of impressions? For example, apparently when you smile a lot and have rapid movements, people take this as if you are an extrovert.</li><li>How does non-verbal behaviour reinforce the nature of a relationship? Apparently men tend to lean forward more and gaze toward the person they are talking with, when the other person so happens to be a female, this behaviour is experienced even more if they are having an intimate conversation, rather than the general boring interpersonal water cooler chit chat.</li><li>Can facial movements be mapped into emotional signals and conversation signals? This is an area greatly studied in deception detection, as there are certain muscles that are expected to be moved when someone is angry, happy, sad etc. Therefore a person might unwillingly move those muscles, and show their true feelings. Or not move them, and therefore give clues as to the fact that they are being deceitful ( See [4]).</li></ul><br />Now the big question is: O.K., So how do social signals help machines understand human emotion?<br />In 2007, a professor and researcher from MIT's Media Lab,Alex Pentland, introduced the notion of “social signal processing”. Which is about applying traditional signal processing techniques to social signals, and use this processing and analysis to predict human social behaviour. For example, his group created a machine that was able to autonomously predict the outcome of a negotiation or of a speed date within its very first minute (see [1] and [2]).<br /><br />The main goal of social signal processing, is to enable analysis of Human behaviour by computers. For this advanced pattern recognition techniques are utilized to automatically interpret complex human behavioural patterns. In the next days, we will talk more about these techniques that are used to interpret human behaviour. Stay tuned! n_n<br /><br />References:<br />1)I. Poggi and F. DÉrrico, "Social signals: A psychological perspective." . Springer Verlag’s Advances in Pattern Recognition series, 2011, pp. 185-225.<br />2)Pentland, A.: Social signal processing. IEEE Signal Process. Mag. 24(4), 108–111 (2007)<br />3)Curhan, J., Pentland, A.: Thin slices of negotiation: predicting outcomes from conversational<br />dynamics within the first five minutes. J. Appl. Psychol. 92, 802–811 (2007)<br />4)Ekman, P., Friesen, W.V.: Nonverbal leakage and clues to deception. Psychiatry 32, 88–106<br />(1969)<br />5) A. A. Salah, M. Pantic, and A. Vinciarelli, "Recent Developments in Social Signal Processing," in Proceedings of the IEEE International Conference on Systems, Man and Cybernetics, 2011, pp. 380-385.Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com0tag:blogger.com,1999:blog-30573458.post-22630877514274591752011-10-14T15:02:00.013-06:002011-10-16T00:02:46.341-06:00Alles Auf Anfang, o la cancion para empezar tu vidaTenia meses que queria compartir esta cancion. Es una cancion alemana de uno de mis grupos favoritos: Wir sind Helden. (Nosotros somos los heores!) . Esta cancion me encanta la melodia, es de esas cancione que me pone de buen humor escucharla, me dan ganas de bailar. Pero lo mejor que tiene esta cancion es la letra. El mensaje que yo entiendo es que debes tomar accion! Si hay cosas que no te agradan de la vida, esta en tus manos cambiarlas. Ponte las pilas. Da todo cada dia! Alles auf anfang! Da todo en este principio que viene! Venga Pumas vamos!<br /><br />Aqui esta la cancion. Enjoy!<br /><iframe width="560" height="315" src="http://www.youtube.com/embed/VtQDWGrX78c" frameborder="0" allowfullscreen></iframe><br /><br /><span style="font-weight: bold;">Version Alemana:</span><br /><br /><div style="border: 5px dashed green;"><br /><b> 23.55: Alles auf Anfang</b><br />Du wirst zahnlos geboren und ohne Zähne gewogen<br />Kriegst sie bis Mitte zwanzig, schon wieder gezogen<br />Bist oh so verschüchtert, verzagt und vernagelt<br />Kein Licht dringt zu dir, so geplagt bist du, sternhageldicht<br />Was dich runterzieht, sind deine schweren Arme<br />Wer schleicht, dem wird leicht kalt, darum schleichst du ins Warme<br />Du nennst es Weltschmerz, ich nenn' es Attitüde<br />Es ist erst fünf vor zwölf und du bist schon so müde<br /><br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br /><br />Nimm deine Zähne, leg sie unter dein Kissen<br />Und sag der Fee du möchtest folgendes wissen:<br />"Warum sinkt mir mein Herz in meine schweren Beine?<br />Ich kann kein Ende sehen von meiner langen Leine"<br />Das was dich so beschwert, das sind die dicken Bären<br />die sie dir aufbinden, du könntest dich beschweren<br />Ob das von Bein haut, das wäre nun zu klären<br />Wenn die kleinlauten, kleinen Leute im Kleinen deutlich lauter wären<br /><br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br /><br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br /><br />Wer "A" sagt muss auch "B" sagen<br />Nach dem ganzen "ABC" fragen<br />Wer "ach" sagt muss auch wehklagen<br />Wer "ja" sagt auch "ach nee" sagen<br /><br />Fühlst du dich mutlos? Fass endlich Mut, los!<br />Fühlst du dich hilflos? Geh' raus und hilf, los!<br />Fühlst du dich machtlos? Geh' raus und mach, los!<br />Fühlst du dich haltlos? Such Halt und lass los!<br /><br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br /><br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br /><br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Fünf vor zwölf, alles auf Anfang"<br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Vier vor zwölf, alles auf Anfang"<br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Drei vor zwölf, alles auf Anfang"<br />Ihr sagt: "Kein Ende in Sicht"<br />Wir sagen: "Zwei, eins, auf die Zwölf"<br /><br /></div><br /><br /><h><span style="font-weight: bold;">Version En Espa~ol!</span></h><br /><div style="border: 5px dashed purple;"><br /><b> 23.55: Listos para empezar!</b><br />Naciste sin dientes y sin ellos te pesaron ,<br />Haz que lleguen hasta tus veinti-tantos<br />Con buena fe arragantelos.<br />Estas tan intimidado, tan fracasado y atrapado,<br />Ningun rayo de luz llega a ti, estas tan molesto<br /><br />Lo que a ti te cansa, son tus brazos pesados,<br />quien anda a hurtillas, a escondidas, estara un poco friolento, por eso entras tu al calor a escondidas.<br />Tu lo llamas "Cansancio de estar vivo", yo lo llamo 'Actitud'<br />Ya son 5 para las doce, y tu ya estas cansado<br /><br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia"<br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br /><br />Toma tus dientes y dejalos debajo de tu cojin,<br />y preguntale al raton de los dientes, todo lo que quieras saber:<br />"Por que mi corazon se cae hasta mis pesadas piernas?<br />No puedo ver el fin desde mi larga linea"<br />Lo que a ti te pesa es que te estan haciendo pendejo,te puedes ir a quejar,<br />talvez sea un problema con la piel de tu pierna, se puede eso aclarar,<br />cuando la gente chiquita y docil se une, es mucho mas fuerte.<br /><br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br /><br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos:"Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br /><br />Quien dice A tiene tambien que decir B<br />Pide por todo el alfabeto<br />Quien dice "ahh": tambien debe empezar a llorar<br />Quien dice "si" debe tambien decir "duh!, obvio!"<br /><br />te sientes sin animos? Vamos animate!<br />Te sientes sin ayuda? Vamos sal a ayudar a la gente!<br />Te sientes impotente? Vamos sal y hazte cargo!<br />Te sientes desorientado? Vamos sal y orientate!<br /><br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br /><br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br /><br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: "Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br /><br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos:"Son las cinco para las doce, vamos a dar el todo en este nuevo dia<br />Ustedes dicen: "No se ve el fin"<br />Nosotros decimos: dos, uno...vamos por el doce!<br /><br /></div><br /><br /><br />Algo que me agrado de hacer esta traduccion, fue que aprendi una frase en aleman nueva:<br />" jmd. einen Bären aufbinden", es como bromear con alguien, yo lo tome como vacilarlo, hacerlo pendejo etc. Se me hace una frase rara porque Bären es oso. O.oLittle Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com5tag:blogger.com,1999:blog-30573458.post-54076824021628270552011-10-06T17:40:00.008-06:002011-10-06T23:31:09.106-06:00El arte de negociar, y la diferencia de generos<a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="https://fbcdn-sphotos-a.akamaihd.net/hphotos-ak-ash4/315011_10150855172550437_814810436_21110139_1499418942_n.jpg"><img style="display:block; margin:0px auto 10px; text-align:center;cursor:pointer; cursor:hand;width: 480px; height: 640px;" src="https://fbcdn-sphotos-a.akamaihd.net/hphotos-ak-ash4/315011_10150855172550437_814810436_21110139_1499418942_n.jpg" border="0" alt="" /></a><br /><br /><br />El dia de hoy mi universidad invito a la autora de Ask For It: How Women Can Use the Power of Negotiation To Get What They Really Want y Women Don’t Ask:The High Cost of Avoiding Negotiation and Positive Strategies for Change, Sara Laschever. Pense en hacer un peque~o post respecto a lo que aprendi en esta platica y compartirlo con mis lectores, porque creo que para muchos (no solo mujeres) el pedir por las cosas es un acto dificil.<br /><br />Dentro de la platica, Sara hablo de como los hombres ven el negociar muy diferente a como lo ven las mujeres. Para los hombres. negociar es algo agradable es como un juego de baseball donde debes tener estrategias. Para las mujeres, el negociar es algo tedioso, algo horrible, como ir al dentista. La autora dijo que esta diferencia de percepcion, radicaba en la diferencia de crianza que existia entre ni~os y ni~as. A las nenas, usualmente se les da juegos que involucran el cuidado de los demas: les dan bebes de juguete, sets de cocina etc. Mientras que a los ni~os, se les dan juegetes donde tienen que explorar su propio ingenio para salir adelante: se les da sets de trenes, donde deben construir rutas y ver como saltar obstaculos etc. A las ni~as tambien se les suele dar tareas diferentes a la de los ni~os. A las ni~as las tareas que se les da son relacionadas con cuidar bebes o a sus herman@s peque~os, ayudar en la cocina. Usualmente todas las tareas en las que se involucran a las nenas hay un adulto supervisando, mientras que a los ni~os, las tareas que se les asigna en el hogar tienen que ver con lavar el coche, quitar la nieve de la acera, arreglar el jardin, sacar la basura etc. Los ni~os reciben menos supervision que las ni~as en las tareas que se les da, y en muchos casos a los ni~os se les paga por el trabajo que ejecutaran: hey te dare 10 pesos si lavas el coche etc. Desde chicos, los ni~os aprenden a negociar las cosas, porque comienzan a decirle a sus padres: Solo 10 pesos? Pero es un auto grande y ademas lo aspirare, dame mejor 15 pesos! Mientras que las ni~as se acostumbran a hacer sus quehaceres por amor. "Por amor cuidare a mis hermanos."<br />Adicionalmente la sociedad, ve mal a las mujeres demandantes, mandonas, e interesadas en el dinero<br />Estas cosas provocan que cuando crezcan, los hombres y las mujeres tengan muy distintos sentimientos respecto al acto de negociar. Esto explica porque, mientras el 65% de los hombres pide un incremento de salario, solo el 12% de las mujeres lo hace.<br />El no negociar o pedir las cosas, hace que uno tenga grandes perdidas. Porque la persona que pidio las cosas, tiene ya un mejor CV que la persona que no pidio nada. La autora hablo de casos, donde los hombres pedian a su Universidad dinero para asisitir a conferencias. La Universidad les daba el dinero y los hombres hacian grandes conexiones por haber podido asistir. Las mujeres, como nunca preguntaron si era posible que la universidad les pagara el viaje, perdian la oportunidad de asisitir a la conferencia y expandir sus horizontes.<br /><br />la autora hablo de verios puntos para mejorar la negociacion. Algunos de ellos son:<br /><ul><li>Asume que TODO es negociable</li><li>Piensa que el mundo es tu ostia ( Tu tesoro). Todo es una oportunidad.</li><li>Vuelvete mas chingona. ( crea conexiones con gente que esta en el poder, estudia una segunda carrera para tener mejor CV, obten diferentes asesores, gente que te puede dar consejos)</li><li>Haz tu investigacion (Obten informacion de cuanto puedes pedir, hay recurso en internet que te muestran salarios promedios de diferentes compa~ias, pregunta con tus amistades.Es importante estar bien informado)</li></ul><br />Por ultimo, algo que dijo la autora que me gusto, es que si aceptas un mal salario, es aceptar que eres chafa. Es como el vino, usualmente si ves un vino barato que cuesta 20 pesos, no esperas mucha calidad de el, en cambio si ves un vino de 200 pesos, es probable que consideres que es de mucha mejor calidad y sabor. Entonces cuando aceptas salarios bajos, estas comunicando algo de ti, estas diciendo que eres el vino de mala calidad de 20 varos. Lo cual no es algo que quieres! Acepta siempre buenos tratos de buena calidad. Tu lo vales! ( Ja comercial loreal ;)<br /><br />Y algo curioso que dijo la autora, es que para poder persuadir a las personas, es importante que la mujer sea amigable. (Esto por lo mismo que se menciono anteriormente, que la sociedad ve mal que la mujer sea mandona y agresiva).<br /><br />Por ultimo, me gustaria escribir sobre el consejo que uno de mis amigos me dio, respecto a pedir cosas: La persona a quien le pediras X cosa, es un adulto que sabe decir NO. Entonces si no peude dartelo, sabe decir NO. no tienes por que preocuparte, no es una situacion incomoda para la otra persona. Y si lo pides, estas mejor que si no lo pides, porque en el peor de los casos, estas donde empezaste.Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com1tag:blogger.com,1999:blog-30573458.post-18639254076320507502011-07-01T22:03:00.000-06:002015-01-10T17:05:37.501-07:00Getting Images laid (pause) over a video in androidThis post is a short tutorial on how to overlay images on video in android. I created this tutorial, after making an android application that plays a video and with certain user interactions displays images on top of the video. This Image-Video effect can also be achieved through action script, but in this tutorial we avoid any extra programming tools and stick to working with the android API.<br />
<br />
Before we begin, we need to giver a quick overview of concepts:<br />
<br />
Most of the user interface components on Android are Views. A View represents a rectangular area on the screen and it is responsible for drawing and event handling. The ImageView class displays an arbitrary image, such as an icon. The VideoView class displays a video file. The ViewGroup class is considered a special view that can contain other views (called children.) This class is the base class for layouts.<br />
<br />
In Android, a layout holds all the elements that appear to the user, and defines where they will be placed. The layout can be declared in an XML file or can be programmatically defined by creating View Objects. A particular type of layout is RelativeLayout. This Class holds the concept that each component in the interface can be described in relation to each other or to its parent.<br />
<br />
The overall idea is that the image on video overlay can be accomplished by using RelativeLayout and placing the VideoView as the first child of the RelativeLayout and the ImageView as the second child. This way in the camera preview, the ImageView will appear to be "on top of" the VideoView.<br />
<br />
The step-by-step instructions are as follows:<br />
<br />
1. In Eclipse, create a Simple Android Project From Scratch (Make sure to have created a main activity).<br />
<br />
2. Under the res folder in your project go to the drawable folder (if you don't have a folder titled "drawable" in res, create it) and add all of the images you plan on working with there.<br />
<br />
3. Add the video or videos you plan on working with to the SD memory card of your android phone. (This can be done by connecting your phone via USB to your PC and on the phone, selecting "notifications", then "USB connected", and in the new window that appears clicking: "Turn on USB storage". After a few seconds a new removable disk should appear on your PC. Copy to it your videos. The video format of my videos was mp4)<br />
<br />
4. Back in your android project in your layout folder, add a new xml file with the name of your choice (for example video_over_image.xml) .In this xml we will define the elements and the layout of our application. For this particular application, we want the following layout:<br />
A text-box in the upper part of the window, where the user types the name of the video they wish to play. The space below the text-box is where the video will be displayed.<br />
The image overlaid on the video will appear on the upper portion of the video. (But it is possible to place it wherever one desires).<br />
Our XML file to accomplish this layout is as follows:<br />
<br />
<div style="border: 2px dashed red;">
<br />
<span style="font-family: courier new;"><?xml version="1.0" encoding="utf-8"?></span><br />
<br />
<span style="font-family: courier new;"><LinearLayout xmlns:android="http://schemas.android.com/apk/res/android"</span><br />
<span style="font-family: courier new;"> android:orientation="vertical"</span><br />
<span style="font-family: courier new;"> android:layout_width="fill_parent"</span><br />
<span style="font-family: courier new;"> android:layout_height="fill_parent"></span><br />
<br />
<span style="font-family: courier new;"> <TextView android:id="@+id/label"</span><br />
<span style="font-family: courier new;"> android:layout_width="fill_parent"</span><br />
<span style="font-family: courier new;"> android:layout_height="wrap_content"</span><br />
<span style="font-family: courier new;"> android:text="Type video name here:" /></span><br />
<span style="font-family: courier new;"> <EditText</span><br />
<span style="font-family: courier new;"> android:id="@+id/edittext"</span><br />
<span style="font-family: courier new;"> android:layout_width="fill_parent"</span><br />
<span style="font-family: courier new;"> android:layout_height="wrap_content"/></span><br />
<br />
<span style="font-family: courier new;"> <RelativeLayout</span><br />
<span style="font-family: courier new;"> xmlns:android="http://schemas.android.com/apk/res/android"</span><br />
<span style="font-family: courier new;"> android:orientation="vertical"</span><br />
<span style="font-family: courier new;"> android:layout_width="fill_parent"</span><br />
<span style="font-family: courier new;"> android:layout_height="fill_parent"></span><br />
<br />
<span style="font-family: courier new;"> <Button android:id="@+id/topBtn"</span><br />
<span style="font-family: courier new;"> android:layout_width="wrap_content"</span><br />
<span style="font-family: courier new;"> android:layout_height="wrap_content"</span><br />
<span style="font-family: courier new;"> android:text="Top"</span><br />
<span style="font-family: courier new;"> android:layout_centerHorizontal="true"></span><br />
<span style="font-family: courier new;"> </Button></span><br />
<br />
<span style="font-family: courier new;"> <VideoView android:id="@+id/surface_view"</span><br />
<span style="font-family: courier new;"> android:layout_width="wrap_content"</span><br />
<span style="font-family: courier new;"> android:layout_height="wrap_content"</span><br />
<span style="font-family: courier new;"> /></span><br />
<br />
<span style="font-family: courier new;"><ImageView android:id="@+id/overlayImage"</span><br />
<span style="font-family: courier new;"> android:layout_width="wrap_content"</span><br />
<span style="font-family: courier new;"> android:layout_height="wrap_content"</span><br />
<br />
<span style="font-family: courier new;"> android:layout_below="@+id/topBtn"</span><br />
<span style="font-family: courier new;"> /></span><br />
<br />
<span style="font-family: courier new;"> </RelativeLayout></span><br />
<br />
<span style="font-family: courier new;"> </LinearLayout></span></div>
<br />
<br />
<br />
An interesting point to notice about this layout is that we added a dummy button to it. Because the video is defined right after this dummy button, the video will completely "cover" the button, so it will not appear on the interface. This button helps in positioning our image; Our image is set relative to this button. In this case, because we sought for the image to appear in the "mid-top" portion of the video the image's layout was set to be below this button. It is also important to note, that the image was declared after the video, because this permits the image to be displayed "on top of" the video.<br />
<br />
<br />
5. In our java activity file in the onCreate method, we now need to establish that we will be using this layout. We also need to establish a listener for the textbox, which after the user has typed the name of the video to play and pressed "enter" will start playing the desired video. Furthermore it is also necessary to establish what images will be overlaid and when that will occur.<br />
To facilitate this example, we will establish that when the user types 1, video A (which should already be on the phone's SD card) will be played and image c1 will be overlaid on the video. Similarly, when the user types 2, video B will be played and image c2 will now be overlaid on the video. We will also add some effects to the image, in specific alpha blending.<br />
In the following, we will present all the code to accomplish this task and subsequently review each part of it:<br />
<br />
<div style="border: 2px dashed red;">
<br />
<br />
<span style="font-family: courier new;">package com.example.android.videooverimage;</span><br />
<br />
<br />
<span style="font-family: courier new;">import android.app.Activity;</span><br />
<span style="font-family: courier new;">import android.media.MediaPlayer;</span><br />
<span style="font-family: courier new;">import android.media.MediaPlayer.OnCompletionListener;</span><br />
<span style="font-family: courier new;">import android.os.Bundle;</span><br />
<span style="font-family: courier new;">import android.util.Log;</span><br />
<span style="font-family: courier new;">import android.widget.MediaController;</span><br />
<span style="font-family: courier new;">import android.widget.VideoView;</span><br />
<span style="font-family: courier new;">import android.net.Uri;</span><br />
<span style="font-family: courier new;">import android.widget.EditText;</span><br />
<span style="font-family: courier new;">import android.view.KeyEvent;</span><br />
<span style="font-family: courier new;">import android.view.View.OnKeyListener;</span><br />
<span style="font-family: courier new;">import android.view.View;</span><br />
<span style="font-family: courier new;">import android.content.res.Resources;</span><br />
<span style="font-family: courier new;">import android.widget.ImageView;</span><br />
<br />
<br />
<br />
<span style="font-family: courier new;">public class VideoOverImageActivity extends Activity </span><br />
<span style="font-family: courier new;">{</span><br />
<br />
<br />
<br />
<span style="font-family: courier new;"> public </span><span style="font-family: courier new;">VideoOverImageActivity</span><span style="font-family: courier new;">() </span><br />
<span style="font-family: courier new;"> {</span><br />
<br />
<span style="font-family: courier new;"> }</span><br />
<br />
<br />
<span style="font-family: courier new;"> public void onCreate(Bundle icicle) </span><br />
<span style="font-family: courier new;"> {</span><br />
<br />
<span style="font-family: courier new;"> super.onCreate(icicle);</span><br />
<span style="font-family: courier new;"> setContentView(R.layout.</span><span style="font-family: courier new;">video_over_image_activity</span><span style="font-family: courier new;">);</span><br />
<span style="font-family: courier new;"> final EditText edittext = (EditText) findViewById(R.id.edittext);</span><br />
<br />
<span style="font-family: courier new;"> edittext.setOnKeyListener(new OnKeyListener() </span><br />
<span style="font-family: courier new;"> {</span><br />
<span style="font-family: courier new;"> public boolean onKey(View v, int keyCode, KeyEvent event) </span><br />
<span style="font-family: courier new;"> {</span><br />
<span style="font-family: courier new;"> // If the event is a key-down event on the "enter" button</span><br />
<span style="font-family: courier new;"> if ((event.getAction() == KeyEvent.ACTION_DOWN) && (keyCode == KeyEvent.KEYCODE_ENTER)) {</span><br />
<span style="font-family: courier new;"> // Perform action on key press</span><br />
<br />
<span style="font-family: courier new;"> int aInt=0;</span><br />
<span style="font-family: courier new;"> try</span><br />
<span style="font-family: courier new;"> {</span><br />
<span style="font-family: courier new;"> aInt = Integer.parseInt(edittext.getText().toString());</span><br />
<span style="font-family: courier new;"> }</span><br />
<span style="font-family: courier new;"> catch (NumberFormatException e)</span><br />
<span style="font-family: courier new;"> {</span><br />
<span style="font-family: courier new;"> Log.e("debug","Error finding Image: "+e.getMessage());</span><br />
<br />
<span style="font-family: courier new;"> }</span><br />
<br />
<span style="font-family: courier new;"> VideoView videoHolder = (VideoView) findViewById(R.id.surface_view);</span><br />
<span style="font-family: courier new;"> MediaController mc=new MediaController(</span><span style="font-family: courier new;">VideoOverImageActivity</span><span style="font-family: courier new;">.this);</span><br />
<span style="font-family: courier new;"> Boolean returnValue=true;</span><br />
<br />
<span style="font-family: courier new;"> switch (aInt) </span><br />
<span style="font-family: courier new;"> {</span><br />
<br />
<span style="font-family: courier new;"> case 1:</span><br />
<br />
<br />
<span style="font-family: courier new;"> startPlaying(videoHolder,mc,"file:///sdcard/video1.mp4",0);</span><br />
<span style="font-family: courier new;"> break;</span><br />
<br />
<span style="font-family: courier new;"> case 2:</span><br />
<span style="font-family: courier new;"> startPlaying(videoHolder,mc,"file:///sdcard/video2.mp4",1);</span><br />
<span style="font-family: courier new;"> break;</span><br />
<br />
<span style="font-family: courier new;"> default: </span><br />
<span style="font-family: courier new;"> returnValue=false; </span><br />
<span style="font-family: courier new;"> break;</span><br />
<br />
<span style="font-family: courier new;"> }</span><br />
<br />
<br />
<span style="font-family: courier new;"> return returnValue;</span><br />
<br />
<br />
<span style="font-family: courier new;"> }</span><br />
<br />
<span style="font-family: courier new;"> return false;</span><br />
<span style="font-family: courier new;"> }</span><br />
<span style="font-family: courier new;"> });</span><br />
<br />
<span style="font-family: courier new;"> }</span><br />
<br />
<span style="font-family: courier new;"> public void startPlaying(VideoView videoHolder,MediaController mc,String nameVideo, int song)</span><br />
<span style="font-family: courier new;"> {</span><br />
<br />
<span style="font-family: courier new;"> Resources res = </span><span style="font-family: courier new;">VideoOverImageActivity</span><span style="font-family: courier new;">.this.getResources();</span><br />
<br />
<br />
<span style="font-family: courier new;"> ImageView image = (ImageView) findViewById(R.id.overlayImage);</span><br />
<br />
<span style="font-family: courier new;"> try</span><br />
<span style="font-family: courier new;"> {</span><br />
<span style="font-family: courier new;"> R.drawable.class.getField("b" + song).getInt(0);</span><br />
<span style="font-family: courier new;"> image.setImageDrawable(res.getDrawable(R.drawable.class.getField("b" + song).getInt(0)));</span><br />
<span style="font-family: courier new;"> image.getDrawable().setAlpha(55);</span><br />
<span style="font-family: courier new;"> }</span><br />
<span style="font-family: courier new;"> catch (Exception e)</span><br />
<span style="font-family: courier new;"> {</span><br />
<span style="font-family: courier new;"> Log.e("debug","Error finding Image: "+e.getMessage());</span><br />
<br />
<span style="font-family: courier new;"> }</span><br />
<br />
<span style="font-family: courier new;"> videoHolder.setMediaController(mc);</span><br />
<span style="font-family: courier new;"> videoHolder.setVideoURI(Uri.parse(nameVideo));</span><br />
<span style="font-family: courier new;"> videoHolder.requestFocus();</span><br />
<span style="font-family: courier new;"> videoHolder.start();</span><br />
<br />
<span style="font-family: courier new;"> videoHolder.setOnCompletionListener(new OnCompletionListener()</span><br />
<span style="font-family: courier new;"> { </span><br />
<span style="font-family: courier new;"> public void onCompletion(MediaPlayer arg0) </span><br />
<span style="font-family: courier new;"> { </span><br />
<span style="font-family: courier new;"> try</span><br />
<span style="font-family: courier new;"> {</span><br />
<span style="font-family: courier new;"> Log.e("debug","MediaPlayer seek to 0...");</span><br />
<span style="font-family: courier new;"> arg0.seekTo(0);</span><br />
<span style="font-family: courier new;"> Log.e("debug","MediaPlayer start...");</span><br />
<span style="font-family: courier new;"> arg0.start();</span><br />
<span style="font-family: courier new;"> Log.e("debug","MediaPlayer started");</span><br />
<span style="font-family: courier new;"> }</span><br />
<br />
<span style="font-family: courier new;"> catch(Exception ex)</span><br />
<span style="font-family: courier new;"> {</span><br />
<span style="font-family: courier new;"> Log.e("debug","MediaPlayer error: "+ex.toString());</span><br />
<span style="font-family: courier new;"> }</span><br />
<span style="font-family: courier new;"> } </span><br />
<span style="font-family: courier new;"> });</span><br />
<br />
<br />
<br />
<span style="font-family: courier new;"> }</span><br />
<br />
<br />
<span style="font-family: courier new;">}</span></div>
Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com4tag:blogger.com,1999:blog-30573458.post-64127034027985881272011-06-30T15:46:00.014-06:002011-09-30T10:35:20.464-06:00Android SDK on Windows for Dummies: A focus on the debuggin part<a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="http://thecellphonejunkie.files.wordpress.com/2010/12/android-2-2-froyo.jpg"><img style="margin: 0px auto 10px; display: block; text-align: center; cursor: pointer; width: 343px; height: 353px;" src="http://thecellphonejunkie.files.wordpress.com/2010/12/android-2-2-froyo.jpg" alt="" border="0" /></a><br />I have recently started developing android apps on Windows 7 (life sucks, everyday I wish I were on Linux, but it is what it is). Today for testing my apps I was given a very unique smart phone from a dubious manufacturer. Since it was not the typical android phone, the normal procedures from : <a href="http://developer.android.com/sdk/win-usb.html" class="external text" title="http://developer.android.com/sdk/win-usb.html" rel="nofollow">http://developer.android.com/sdk/win-usb.html#WinUsbDriver</a> did not seem to work :( I followed all of Google's instructions. But when I tried to update the driver I encountered a few problems: windows asked me where to search for the driver software, and I selected: "Browse my Computer for driver software", then clicked "Browse" and Explore to C:\Android\usb_driver. I also checked the "Include Subfolders", clicked Next and... I got the following error message: "Windows was unable to install your android phone"<br /><br />After hours and hours of working around it, I finally found a solution to the problem and thought I'd share it, so people can avoid some of the pitfalls I encountered and the installation will hopefully not be as challenging as it was for me.<br /><br />Basically what worked for me was to download PdaNet for android from : <a href="http://www.junefabrics.com/android/download.php">http://www.junefabrics.com/android/download.php</a> . I installed PdaNet with the phone connected to the PC and android was up and running (and not suspended).<br /><br />PdaNet is technically a tool for supplying Internet access to an unconnected device from a device (such as a mobile phone) which does have Internet access. I believe PdaNet is useful in this case, because it automatically sets up all of the environment for having communication between the computer and the phone.<br /><br />Once PdaNet has been successfully installed, I ran from a windows command propt "adb.exe" and "fastboot.exe". Now, when I ran the latter, I received a message stating that a .dll file was not found, I search for that file, and added its location to my path.<br />Here it might be important to state that adb.exe is the "android debug bridge", a tool that can deal with the emulator or the device. Fastboot on the other hand, is a diagnostic protocol used primarily to modify the flash file system in Android smart phones from another computer over a USB connection.<br /><br />With this,I had communication with my android phone!<br /><br />One can test it out, by typing in a command propt: <span style="font-weight: bold;">adb devices</span> and obtain a list of connected devices, including the phone. With that communication between our device and our computer is achieved, so running and testing our application is now a cinch.<br />happy hacking :)Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com1tag:blogger.com,1999:blog-30573458.post-34281719218551363722011-05-16T23:19:00.004-06:002011-05-16T23:34:31.483-06:00LDA is not Ladies Ditching ApesI have recently been working on Topic Modelling and thought I'd do a brief tutorial on how to automatically divide a text into a series of relevant topics.<br /><br />Before we dive into our coding, let's give a brief overview of the topic so we are all on the same page:<br /><br />Topic Modelling is all about automatically finding the thematic structure of a document or a series of documents.<br /><br />Topic modelling specifies a probabilistic method through which documents can be created. Initially a distribution over topics is selected. For example, the topics of Love and Mexico could be chosen, but each assigned a certain weight or probability. If it was sought for the article to have a greater political inclination, Mexico would be assigned a greater weight than Love. Whereas if the purpose was to write a romantic novel, Love would have a much higher weight or probability assigned than Mexico. Once the topics along with their corresponding probabilities have been assigned, a topic is chosen randomly according to the distribution, and a word from that topic is drawn. This process of randomly choosing a word from a topic is done iteratively until the system has finished "writing" the article.<br /><br />Besides creating documents automatically ( Hasta la vista estudiantes de Literatura :P) Topic Modelling can also infer the set of topics responsible for generating a collection of documents.<br />We care about Topic Modelling because it can enhance search in large archives of texts, it also permits for better similarity measures: given two documents exactly how similar are they?<br /><br />Different algorithms exist for finding the thematic structure of a document. Today we will focus on one particular algorithm called Latent Dirichlet Allocation (LDA). Which is a "...generative probabilistic model for text corpora...".<br />The intuition behind LDA is that a document is conformed of a series of different topics, and each topic is a probability distribution over words. Each document is a random mixture of corpus-wide topics, where each word of a document is drawn from one of these topics. LDA intents to infer how the documents are divided according to these topics, and what the topics are. The only information LDA has, are the documents.<br /><br />In the following, we will use THE FORMAL notation of LDA, (mathematical style!) to make things a bit more clearer:<br />P(z) denotes the topic distribution z of a particular document. P( w | z ) is the probability distribution of words w given topic z. LDA assumes each word wi in a document (where the index refers to the ith word token) is generated by first sampling a topic from the topic distribution, then choosing a word from the topic-word distribution. We write P( zi = j ) as the probability that the jth topic was sampled for the ith word token and P( wi | zi = j ) as the probability of word wi under topic j.<br /><br />LDA assumes that the topics present a Dirichlet distribuition, i.e. the mixture weights θ are generated by a Dirichlet prior on θ. Each topic is modelled as a multinomial distribution<br />over words.<br />Hopefully this brief overview will allow us to have some Python coding fun for our next post!<br /><br /><a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="http://www.quarterinchpublishing.com/fabric/images/fabric_ninos_words_green.jpg"><img style="display: block; margin: 0px auto 10px; text-align: center; cursor: pointer; width: 500px; height: 634px;" src="http://www.quarterinchpublishing.com/fabric/images/fabric_ninos_words_green.jpg" alt="" border="0" /></a>Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com6tag:blogger.com,1999:blog-30573458.post-78976118380037250702011-03-05T16:12:00.067-07:002011-03-14T11:56:27.957-06:00How to use Machine Learning to boost up your Parallel ComputingIn the past, programmers would find reassurance to their problems of running extremely large programs in the yearly speed up of computer processors. Every year or so, the speed of computers would double and a faster computer would be in the market which would be able to rapidly execute their large sized code. But, in today's world this is not the case, analysing the GHz speed of the processors in a desktop computer of 2 years ago, versus the speed new desktop computers present, proves that it has barely, if any, increased, this mainly due to the fact that it is very difficult to build reliable low-cost processors that run significantly faster. This is the reason why, the solution to running faster code has focused on doubling the number of processors that exist on a single chip. Actually researchers believe, that in the following years we will have systems, which present twice the number of cores with every new technology generation. <br /> <br />It is important to note, that these multi-processors have laid the road for using parallel computing, in which a large program can be divided into smaller programs, each of which is then assigned to a processor with shared or independent resources. Parallelism is what is generally used today, for providing Performance improvement. <br /><br />This approach although generally highly functional, has shown in some cases to degrade the performance considerably! The problem is that the scheduling of parallel jobs is a very complicated task which is highly dependent on a series of different factors: the workload, the blocking algorithm being utilised, the local operating system, the parallel programming language and the machine architecture. Expert humans are who tend to make the design specifications for these highly complicated tasks,and as a result they tend to be somewhat rigid and unsophisticated [1].<br /><br />Because of this, machine learning techniques have in recent years provided a solution to this problem. Machine learning is a field which intends to build computer systems that automatically improve with experience. Researchers have been applying machine learning algorithms to problems of resource allocation, scheduling, load balancing, and design space exploration, among other things. <br /><br />Such is the work done, in Cost-Aware Parallel Workload Allocation Approach based on Machine Learning Techniques. Here the authors tackle the problem of finding adequate workload allocations in a cost-aware manner, by learning from training examples how to allocate parallel workload among Java threads.<br />Given a program, their system computes its feature vectors, and utilising a nearest neighbour approach finds from the training examples, the best parallel scheme for this new program.<br /><br />One may initially wonder, what type of training examples were utilised for this problem and how were they generated? <br />The training examples came from a series of programs coded in java, which presented different for loops. From each for loop its corresponding feature vector was calculated along with its associated label, this conformed each training example. In this case, the feature vector corresponded simply to the workload the for loop presented, and the label to the optimal number of threads that should be utilised with that specific workload.<br />The programs which were utilised for the training examples were manually selected, each had the purpose of bringing a certain workload variety to the training pool.The labels were set by an automatic program, which tested each workload (loop description) with a different number of threads, and then calculated what was the optimal thread number required for that specific workload to achieve optimal performance. It is important to note, that in this approach the computation cost of calculating the feature vectors was diminished by calculating an implicit estimate of the workload, the features which conformed the workload were: 1) loop depth; 2) loop size; 3) number of arrays used; 4) number of statements within the loop body, and 5) number of array references.<br /><br />Since not all program features play an equal role in workload estimation different weights were assigned to different features during classification, with higher weights given to feature 1), 2) and 4). Within the paper it was not clearly explained how the values of these weights were assigned or what their values were. It might have been adequate to also utilise a learning algorithm which was capable of finding the most adequate weights given a certain training example, because it might be the case that under certain conditions a feature might be weaker for classification than an other, and therefore other weights need to be utilised. A broader explanation on the weight manner could have provided more insight and restricted this speculation, but it is interesting to ponder none the less.<br /><br />On the other hand, in this example the authors opted for a supervised learning approach, where each training example that was handed to the system was manually selected and labelled. This is clearly a tedious task to do, and at times may not be the most optimal, since manually finding which examples provide more information for the learning process in comparison to other possible training examples is difficult and non-trivial. One therefore wonders if an unsupervised learning algorithm could have provided better results. In this type of approach, the machine can be "thrown into the wild" and through observations discover previously unknown structures or relationships between instances or their components, this could eliminate the problems mentioned previously, but has the shortcoming that if the learning phase is done online the algorithm might take much longer than if an instance based learning approach had been taken.<br /><br /><br /><br />Furthermore it does not seem that their approach accounts for any long-term consequences. Each decision within a for loop was done independently of what had been decided for the other for loops within the program, this might mean, the decision to use X amount of threads for that workload might be locally optimal but not globally optimal, this could in the long run deteriorate the performance. This situation seems to suggest that for this problem instead of using instant based learning, a better approach would have been to use Reinforcement Learning. In Reinforcement Learning, the machine is not told which actions to take, but rather must infer them, by analysing what yields the best reward. The following figure presents an overview of how reinforcement learning works<br /><br /><a onblur="try {parent.deselectBloggerImageGracefully();} catch(e) {}" href="http://2.bp.blogspot.com/-4UuyiC8o__E/TX5WVVmAtSI/AAAAAAAABCA/BaMIAMei2pA/s1600/x2.png"><img style="display:block; margin:0px auto 10px; text-align:center;cursor:pointer; cursor:hand;width: 320px; height: 240px;" src="http://2.bp.blogspot.com/-4UuyiC8o__E/TX5WVVmAtSI/AAAAAAAABCA/BaMIAMei2pA/s320/x2.png" border="0" alt=""id="BLOGGER_PHOTO_ID_5583995512559809826" /></a><br />For this particular case, if they authors had used this other learning method, it could have been possible for the machine to analyse the program as a whole, and decide then what the best long-term thread allocation for each workload would be. In specific, the machine would have interacted with the "environment" (in this case the supplied java program) over a discrete set of time steps. In each step the machine would have "sensed" the environment's current state (which would match the number of threads being used in each for loop) and executed an action(an action would correspond to assigning or removing more threads to certain for loops). This action would modify the environment’s state (which the machine could sense in the next time step) and produced an immediate reward (The reward would be the overall performance obtained for that particular thread assignation).<br /> The machine’s objective would be to maximise its long-term cumulative reward by learning an optimal policy that maps states to actions. <br />In its most basic form, Reinforcement Learning brings a knowledge-free trial-and-error methodology in which the machine intents various actions in numerous system states, and learns from the consequences of each action. <br />From this, it is clear that the advantage of using this learning method is that no explicit model of either the computing system being managed or of the external process that generates workload or traffic are necessary.Additionally, Reinforcement Learning is capable of treating dynamical phenomena in the environment, as mentioned before, it can analyse how current decisions may have delayed consequences in both future rewards and future observed states.<br /><br />Now, while this can sound very promising, it is necessary to also take into consideration, the challenges which Reinforcement Learning faces in the real world. Firstly, Reinforcement Learning can suffer from poor scalability in large state spaces, furthermore in times the performance obtained during online training can be below average, due to the lack of domain knowledge or good heuristics. In addition, because reinforcement learning procedures need to include "exploration" of actions, the selection of actions can be exceedingly costly to implement in a live system. This is the reason why, many modern applications that utilise reinforcement learning in order to address the above practical limitations,take a hybrid approach. Such an example, is the work done in A Hybrid Reinforcement Learning Approach to Autonomic Resource Allocation. Here the authors propose for the machine to have an offline training phase. They suggest that given enough training examples which follow a certain optimisation policy , the learner (machine) using reinforcement learning will be able to converge to the correct value function, it will be able to find a new policy which greedily maximises the value function and is able to improve the original policy that was given. In this form, the poor performance that is obtained by using live online training is avoided. Another benefit of their method is that multiple iterations can be done: Through training a new policy, which is the improved version of the original policy, is obtained. This improved policy can then be feed into the system again, acting as the original non-optimal policy, with this second policy a second data set is collected, which can then be used to train a further improved policy. This enables the possibility of running the algorithm iteratively till a desired "reward" is obtained. <br /><br />It was mentioned before that reinforcement learning, presents the problem of having expensive exploration of actions, the authors tackled this problem by replacing the generally used lookup table for representing the value function with a nonlinear function approximator, in particular a neural network. A function approximator provides a solution to the mentioned issue, because it is mechanism for generalising training experience across states, therefore it is no longer necessary to visit every state in the state space. It also allows for generalisation across actions, so that the need for exploratory off-policy actions is also greatly reduced.<br /><br />Their hybrid Reinforcement Learning approach was tested on realistic prototype Data Center, which dynamically allocates servers among multiple web applications so as to maximise the expected sum of SLA (service level agreement) payments in each application.<br /><br />Although their proposed solution resolves most of the problems encountered with reinforcement learning, we can observe an aspect of their work, that might call for improvement: In their algorithm with each iteration, the model of the system is modified. They always assume the model "learned" from the use of a certain set of policies can never be applied to a set conformed of other policies. The authors never explored if this is always the case, could a model learned with certain policies still be valid under other policies which hold a degree of similarity to the original policies, or is it always necessary to learn from scratch the model, as a result of changes to an active set of policies? <br />Additionally, the authors utilised a neural network for finding the states to explore, and although this did solve the exploratory problem mentioned before, because the neural network has hidden states it is not possible to determine beforehand exactly how many states will be explored given the current used policies, knowing beforehand this number could improve computational costs as better planning can be done. In the work done in "An Adaptive Reinforcement Learning Approach to Policy-driven Automatic Management", the authors addressed this problems and show how a Reinforcement Learning Model can be adapted to accommodate this.<br />The authors analysed how previously learned information about the use of policies can be effectively used in a new scenario. For this, they consider policy modifications as well as the amount of time used to form the model before the changes. Similarly to the work in Hybrid Reinforcement Learning Approach to Autonomic Resource Allocation, a state transition model, which uses a set of active expectation policies is defined, but in difference to Hybrid Reinforcement Learning Approach to Autonomic Resource Allocation, instead of using a neural network, the authors capture the management system's behaviour through a state-transition graph, what their system is lacking and could be beneficial in the future is mapping directly how changes in policies effect the state-transition modelsLittle Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com1tag:blogger.com,1999:blog-30573458.post-19901423770989271152011-01-24T23:38:00.009-07:002011-01-25T00:38:36.320-07:00Soñando con Tacos en la ciudad de las estrellas de cine rodeada de angeles<span style="color: rgb(153, 51, 153);">(Este post es dedicado a mi lectora favorita!) </span><br />Mi<a href="http://ihatecomicsans.com.mx/"> lectora favorita</a>, (ya que parece ser la unica que tengo..jajaja :P) Me recomendo ayer una cancion alemana ochentera. En general odio la musica ochentera, I'm all about the sixties man! Pero dado que tenia un estilo peculiar y fue recomendada por mi lectora favorita, decidi hacer un post de Musica Alemana al alcanze Mexicano!<br />Mi traduccion de D.A.F. KEBAB TRäUME!<br /><br /><span style="font-weight: bold;">Version Alemana:</span><br /><br /><div style="border: 2px dashed red;"><br />Kebabträume in der Mauerstadt,<br />Türk-Kültür hinter Stacheldraht<br />Neu-Izmir ist in der DDR,<br />Atatürk der neue Herr.<br />Miliyet für die Sowjetunion,<br />in jeder Imbißstube ein Spion.<br />Im ZK Agent aus Türkei,<br />Deutschland, Deutschland, alles ist vorbei.<br /><br />Kebabträume..<br /><br />Miliyet...<br /><br />Kebabträume...<br /><br />Miliyet...<br /><br />Wir sind die Türken von morgen.<br />Wir sind die Türken von morgen..<br /></div><br /><br /><h><span style="font-weight: bold;">Version En Espa~ol!</span></h><br /><div style="border: 2px dashed blue;"><br />Sue~os de Kebabs en la ciudad del Muro <span style="color: rgb(102, 102, 102);">(Los Kebabs son un platillo tipico turco, usualmente llamado por ellos Döner kebab, la ciudad del Muro se podria referir a Berlin. )</span><br />Cultura turca atras de ese alambre de puas.<br />La nueva capital Turca esta en el este de alemania<br />Atatürk el nuevo Se~or <span style="color: rgb(102, 102, 102);">( Atatürk fue el primer presidente de Turquia!)</span><br />"nacionalidad" para la Union Sovietica. <span style="color: rgb(102, 102, 102);">( La palabra Miliyet no esta en aleman, sino en turco y significa Nacionalidad)</span><br />en cada cafeteria un espia<br />La administracion de los partidos comunitas regidos por alguien de Turquia. <span style="color: rgb(102, 102, 102);">( En la cancion usan la abreviacion ZK que es Zetralkomitee, el cual representaba el cuerpo administrativo de los partidos comunistas en Alemania- De acuerdo a Wikipedia)</span><br />Alemania Alemania, Todos esta perdido.<br />Sue~os con Kebabs...<br />Nacionalismo (en turco)<br />Sue~os con Kebabs...<br />Nacionalismo (en turco)<br />Nosotros somos los Turcos de Ma~ana,<br />Nosotros somos los Turcos de Ma~ana....<br /></div><br /><br />...wow...debo admitir que ME ENCANTO hacer esta traducccion! Muchas gracias a quien la recomendo!<br />No solo me sirivio para recordar el aleman, sino para aprender un poco de historia.<br /><br />Considero que es duro como se refieren los turcos que viven en Alemania a ALemania: " Todo esta perdido." No se si yo me podria atrever a decir algo similar de un pais viviendo alli.<br />Muchos Mexicanos viven en EUA, pero no se si piensen o canten: EUA todo esta perdido...EUA todo esta perdido. Es una cancion muy nacionalista turca, que hace menos a la cultura alemana. Los demas que opinan?<br /><br />Creo que es dificil ser extranjero en Alemania, se que en los trenes los policias tienen derecho a interrogar y pedir boletos a los que vean sospechosos y usualmente la selecion se hace de modo racial. Entonces ha de ser incomodo, no tener los ojos claros y el pelo rubio y verse como un tipico aleman. Talvez de alli viene ese sentimiento de enojo hacia Alemania y decirle que esta acabado, que quienes tienen el poder son ellos.<br />Alguien mas siente que la cancion es extremadamnete agresiva hacia los alemanes?<br /><br />A veces pienso que si me gusta gritarle a los extranjeros el amor que tengo por Mexico, por nuestros tacos al pastor, la barbacoa, los corridos, los sones jaroches, por todas las cositas que son Mexico. Pero no se si me iria al extremo de decirles que su pais esta terminado. Se que por ejemplo, varios federales de EU han matado de modo violenta a la juventud mexicana. Pero aun no siento en mi sangre, tanto odio para cantarles que su pais ya cayo, ya termino.<br />Los Mexicanos que opinan?<br />Sue~o con Tacos..Sue~o con Tacos en la ciudad llena de estrellas de cine y rodeada de Angeles...<br /><iframe src="http://www.youtube.com/embed/El-UTSMl5TU?fs=1" allowfullscreen="" frameborder="0" height="344" width="425"></iframe>Little Saiphhttp://www.blogger.com/profile/11964246294373530295noreply@blogger.com7