<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Vinicius Torres de Macedo on Medium]]></title>
        <description><![CDATA[Stories by Vinicius Torres de Macedo on Medium]]></description>
        <link>https://medium.com/@vtmacedo?source=rss-48b407ffbdb6------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*UvOMeyX2cIo8_-mwGDzQRg.jpeg</url>
            <title>Stories by Vinicius Torres de Macedo on Medium</title>
            <link>https://medium.com/@vtmacedo?source=rss-48b407ffbdb6------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Sat, 21 Nov 2026 18:46:39 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@vtmacedo/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[If you want to make everyone happy, don’t be leader. Sell ice cream]]></title>
            <link>https://medium.com/@vtmacedo/if-you-want-to-make-everyone-happy-dont-be-leader-sell-ice-cream-c8866e707747?source=rss-48b407ffbdb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/c8866e707747</guid>
            <category><![CDATA[collaboration]]></category>
            <category><![CDATA[leadership]]></category>
            <category><![CDATA[teamwork]]></category>
            <category><![CDATA[communication]]></category>
            <category><![CDATA[strategy]]></category>
            <dc:creator><![CDATA[Vinicius Torres de Macedo]]></dc:creator>
            <pubDate>Thu, 20 Jul 2023 19:36:45 GMT</pubDate>
            <atom:updated>2023-07-20T19:36:45.942Z</atom:updated>
            <content:encoded><![CDATA[<p>Steve Jobs is the author of this famous quote, in this post, let’s try to understand what he was trying to say with it. When you ask any group of managers, “What do effective leaders do?” you’ll receive a barrage of answers. They may mention leaders strategizing, motivating, creating a mission, and building a culture. However, when you ask the same group, “What should leaders do?” experienced individuals will have only one answer: the primary job of a leader is to <strong>generate results.</strong></p><p>Despite this consensus, the mystery of what leaders can and should do to promote optimal team performance has persisted. In recent years, this mystery has fueled the growth of a once-nascent industry. Thousands of “leadership experts” have made careers out of testing and training executives with the intention of shaping business men and women capable of turning ambitious objectives, whether strategic, financial, organizational, or all three, into reality.</p><h3><strong>The Six Leadership Styles</strong></h3><p>Imagine the Formula 1 tires as a collection of tools carried by a team leader in a professional race. During a race, the team leader, together with the team’s strategists and engineers, decides and uses each tire based on the specific requirements of the track and the racing conditions. Sometimes they need to make careful decisions about which tire to use, but in general, the selection process becomes automatic. The team leader swiftly assesses the racing challenge ahead, makes the right tire choice, and skillfully puts it to use. Highly impactful leaders operate in a similar manner.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*AOIQddPK00y1ppep.png" /></figure><p>In practice, many managers overlook the significant impact of organizational climate on overall results, despite its influence on approximately one-third of financial performance. The organizational climate, in turn, is heavily influenced by the leadership style employed by managers. This includes how they inspire and motivate their subordinates, gather and utilize information, make decisions, manage change initiatives, and handle crises. There exist six fundamental leadership styles, each rooted in different aspects of emotional intelligence, and each functioning optimally in specific situations while exerting distinct effects on the organizational climate. In the book <em>“HBR’s 10 Must Reads on Managing People”</em> specifically in the article <em>“Leadership That Gets Results”</em> by Daniel Goleman, the author explores these leadership styles through insightful quotes, aiming to provide a comprehensive understanding of each one. Let us delve into the exploration of these styles.</p><ul><li><strong>Coercive Style</strong>: <em>“Do what I say”</em></li></ul><p>While this approach can prove highly effective in recovery situations, natural disasters, or when dealing with troublesome employees, it tends to inhibit organizational flexibility and dampen employee motivation in most other circumstances.</p><ul><li><strong>Visionary Style</strong>:<em> “Come with me”</em></li></ul><p>Leaders employing this approach declare the overall goal and provide individuals with the freedom to choose their own means to achieve it. This style works particularly well when a business is adrift, but it may be less effective when the leader is working with a team of experts who possess greater experience than themselves.</p><ul><li><strong>Affiliative Style</strong>: <em>“People first.”</em></li></ul><p>This style proves particularly useful in building team harmony and boosting morale. However, its praise-focused approach can allow poor performance to go unaddressed. Additionally, affiliative leaders often fail to provide clear guidance, leaving employees undecided.</p><ul><li><strong>Democratic Style</strong>: <em>“What do you think?”</em></li></ul><p>While this style does foster organizational flexibility, accountability, and the generation of new ideas by giving employees a voice in decision-making, its impact on the organizational climate is not as strong as one might expect. It can sometimes lead to endless meetings and employees feeling a lack of clear leadership.</p><ul><li><strong>Pacesetting Style</strong>: <em>“Do what I do, now”</em></li></ul><p>Leaders who set high-performance standards and lead by example can have a highly positive impact on self-motivated and highly competent employees. However, other team members often feel overwhelmed by the demands for excellence and may resent the leader’s tendency to take over.</p><ul><li><strong>Coaching Style</strong>: “Try this”</li></ul><p>This style primarily focuses on personal development rather than immediate task-related work. It works well when employees are self-aware of their weaknesses and motivated to improve. However, it falls short when employees resist changing their own methods and are unwilling to embrace personal growth.</p><p>The more leadership styles a leader masters, the better, especially when they can switch between them according to the circumstances. This fosters an excellent organizational climate and optimizes business performance.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/0*aUtlGMgl_f86421h" /><figcaption>Formula 1 tires types</figcaption></figure><p>In conclusion, effective leadership is a complex art that requires a deep understanding of different leadership styles and their implications. Just like a skilled Formula 1 team leader who must make strategic tire changes during a race, successful leaders must be adaptable and capable of employing various leadership styles based on the specific demands of each situation. However, much like Formula 1 pilots who may not always appreciate the tire changes made by their team leaders due to their personal preferences or past experiences, employees in organizations may also have mixed reactions to certain leadership styles.</p><p>In the dynamic world of business, successful leaders are those who master multiple leadership styles and possess emotional intelligence competencies to navigate diverse challenges. They must be willing to acknowledge and learn from mistakes, just like innovators in Formula 1 quickly adapt their tire choices during races. By combining the right leadership approach with situational awareness and a willingness to embrace change, leaders can create an empowering organizational climate and drive optimal team performance.</p><p>In summary, effective leaders understand that a cohesive and motivated team is the key to sustained success in the ever-evolving world of technology and business. Ayrton Senna, the legendary Brazilian Formula 1 driver, exemplified this philosophy with his famous quote, <em>“I have no idols. I admire work, dedication, and competence.” </em>Senna’s words remind us that true leadership goes beyond individual brilliance, it stems from fostering a team culture built on hard work, dedication, and a relentless pursuit of excellence. In the same way, impactful leaders in any field must inspire and empower their teams, recognizing that success is a collective effort and a reflection of the team’s shared dedication to achieving greatness.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/997/0*EtgbIVNz0eSufGU-.jpg" /><figcaption>Ayrton Senna Podium</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c8866e707747" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[An artificial inteligence approach to the autism spectrum quotient (AQ-10) test.]]></title>
            <link>https://medium.com/@vtmacedo/an-artificial-inteligence-approach-to-the-autism-spectrum-quotient-aq-10-test-20e5b997a7fd?source=rss-48b407ffbdb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/20e5b997a7fd</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[autism-spectrum-disorder]]></category>
            <category><![CDATA[data-augmentation]]></category>
            <dc:creator><![CDATA[Vinicius Torres de Macedo]]></dc:creator>
            <pubDate>Wed, 10 May 2023 18:22:44 GMT</pubDate>
            <atom:updated>2023-05-10T20:29:37.872Z</atom:updated>
            <content:encoded><![CDATA[<p>Autism Spectrum Disorder (ASD) is a complex neurodevelopmental disorder that affects communication, social interaction, and behavior. Although much is still unknown about its causes, research has explored factors that may contribute to its development, such as genetic predisposition, diet, exposure to chemicals, vitamin D and folic acid deficiency, prematurity, and advanced parental age.</p><p>Diagnosing such a diverse disorder that presents in various ways is a challenge. To do so, a rigorous clinical analysis is necessary, often involving professionals such as speech therapists, therapists, doctors, and social workers. Among the various available diagnostic methods, we will focus on one in particular: The Autism Spectrum Quotient (AQ-10).</p><p>The AQ-10 is a questionnaire composed of 10 short multiple-choice questions aimed at assisting in the diagnosis of individuals suspected of having ASD in different age ranges. The proposed project aims to apply an artificial intelligence model to perform a pre-diagnosis of ASD based on the responses to the AQ-10. Additionally, the project seeks to identify which questions from the questionnaire are most relevant for classification. The main objective is to improve both the efficiency and accuracy of diagnosis, with the aim of facilitating access to appropriate interventions for individuals suspected of having ASD.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/0*2VaUN7eNmKGjzAuo" /><figcaption>AQ-10 questionnaire for adults</figcaption></figure><p>The database used in this study was collected through a form filled out by over 700 people, containing responses from the AQ-10 questionnaire and can be reached <a href="https://archive-beta.ics.uci.edu/dataset/426/autism+screening+adult">here</a>.</p><p>Before building an artificial intelligence model, a careful analysis of the database was conducted to identify possible trends, biases, or prejudices. This step was crucial to ensure the accuracy and reliability of the AI model, as it is known that introducing inaccuracies in the data can lead to inaccurate results. Therefore, it is essential to protect the AI model from data that may compromise its accuracy.</p><p>During the dataset analysis, we identified the presence of trends in different columns, such as the country of residence and ethnicity of the interviewees. In addition, we also observed an imbalance in the columns of the form and diagnosis results, as well as a high incidence of false negatives, which represent a serious medical issue to be addressed with caution.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*snjtBirkQZ58vR2aUGTHxA.png" /><figcaption>Quantity of some attributes in the dataset.</figcaption></figure><p>To work with the data, we removed rows that had null values and used the Data Augmentation technique on the class of people who were positively classified as autistic according to the AQ-10 result. Before this process, there was a 6:1 ratio (positive and negative) among those classified. After applying the technique, this ratio became 1:2 (positive and negative) among those classified, which helped balance the dataset.</p><p>Furthermore, to work with the model, we decided to exclude columns referring to gender, ethnicity, country of origin, and others that we considered irrelevant to the dataset work, as this information would not significantly influence the final result. With these steps taken, we are more confident in the quality of the dataset and ready to build an AI model with greater accuracy and efficacy.</p><p>The choice of the Decision Tree model over other more advanced artificial intelligence models is due to a variety of reasons:</p><ul><li>Firstly, its simplicity is its strength.</li><li>Completely deterministic model, meaning it produces precise and predictable results, which is an advantage when dealing with medical and health issues</li><li>The Decision Tree model is easy to understand and interpret, making it more accessible to researchers who do not have a deep knowledge of machine learning algorithms.</li></ul><p>This means that we can trust the results generated by the model without fear that uncertainty will affect the accuracy of the analysis.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*EPTJA-Y1QVeNKIUf.png" /><figcaption>Decision tree example.</figcaption></figure><p>Finally, this model is one of the few that can answer our research question. With the ability to track the results given by this model, we can extract nuances and insights from the paths taken by the model for a given condition or response, which allows us to better understand the criteria used by the model to make the classification. With this information, we can further improve the model and increase its accuracy.</p><p>Going back to the model, we split the dataset into 20 and 80 percent for testing and training respectively. In addition, we used the GridSearchCV technique to discover its best parameters.</p><pre>dt_parameters = {<br>    &quot;clf__criterion&quot;: [&quot;gini&quot;, &quot;entropy&quot;, &quot;log_loss&quot;],<br>    &quot;clf__splitter&quot;: [&quot;best&quot;, &quot;random&quot;],<br>    &quot;clf__max_depth&quot;: range(3, 50),<br>    &quot;clf__min_samples_split&quot;: range(2, 10)<br>}<br><br>dt_gridsearch = search(DecisionTreeClassifier(random_state=random_seed), dt_parameters)<br>dt_best_score = dt_gridsearch.best_score_<br>dt_best_params = dt_gridsearch.best_params_<br><br>print(f&quot;Best score on the decision tree otimization: {dt_best_score}&quot;)<br>print(&quot;Best params:&quot;)<br>for key, value in dt_best_params.items():<br>    print(f&quot;{key}: {value}&quot;)</pre><pre>Best score on the decision tree otimization: 0.7864386317907444<br>Best params:<br>clf__criterion: entropy<br>clf__max_depth: 12<br>clf__min_samples_split: 2<br>clf__splitter: best</pre><p>Expanding a little more on the accuracy rates obtained, we can notice that our model has a considerably better accuracy than the one presented by the dataset. The most important result was the reduction of the false negatives rate.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/646/1*sOsBtqxvMzyLKzF5EQkweg.png" /></figure><p>The following questions from the questionnaire were considered the most important :</p><ul><li>[A2] I usually concentrate more on the whole picture, rather than the small details.</li><li>[A3] I find it easy to do more than one thing at once.</li><li>[A7] When I’m reading a story, I find it difficult to work out the characters’ intentions.</li></ul><p>It is worth noting that we only considered the questions that appeared positively and negatively in the top 15.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1009/1*P_eCYHR7yirnYaNVGpF15Q.png" /><figcaption>Dataset feature relevance.</figcaption></figure><p>There are many possibilities for further exploration with this type of work, here are some examples:</p><ul><li>Integration of the questionnaire with the intelligence model, aiming at continuous improvement.</li><li>The model’s classification can be taken into account within a context of medical screening or diagnostic support.</li><li>Given the patient’s responses, the model’s screening (“path”) followed can be an interesting input for the healthcare professional’s judgment, thus providing insights for the correct diagnosis.</li></ul><p>This work was developed during the discipline of medical data manipulation at the UFPE Informatics Center. I would like to thank my squad formed by <a href="https://medium.com/u/eb7d3e462aeb">José Bezerra</a>, <a href="https://medium.com/u/7289100a063">André Luiz</a>, Pedro Milet, Otávio Augusto and Victor Souza who contributed to the success of the work.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=20e5b997a7fd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Python Serverless Frameworks, which one should I choose? [Chalice, Zappa, Serverless]]]></title>
            <link>https://medium.com/@vtmacedo/python-serverless-frameworks-which-one-should-i-choose-chalice-zappa-serverless-e090bab1757e?source=rss-48b407ffbdb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/e090bab1757e</guid>
            <category><![CDATA[serverless-framework]]></category>
            <category><![CDATA[chalice]]></category>
            <category><![CDATA[serverless]]></category>
            <category><![CDATA[zappa]]></category>
            <category><![CDATA[lambda-function]]></category>
            <dc:creator><![CDATA[Vinicius Torres de Macedo]]></dc:creator>
            <pubDate>Thu, 30 Mar 2023 19:32:18 GMT</pubDate>
            <atom:updated>2023-03-30T20:26:15.838Z</atom:updated>
            <content:encoded><![CDATA[<p>A research by <a href="https://www.datadoghq.com/state-of-serverless-2021/"><strong>Datadog</strong></a> showed that the most popular Lambda runtime, especially in large environments is Python, which has 58% of all users. That’s why, I want to talk about some Python Serverless Frameworks that I tried to use to deploy a lambda function working as a ML model. We will cover 3 technologies in this article: Chalice, Zappa and Serverless.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*Rn-jyIjJa-4ktM70.jpeg" /></figure><p>The first one is a Python Microframework for AWS that allows you to deploy apps that use Amazon API Gateway and AWS Lambda. But, what are the advantages in using a serverless framework? As its name says, you don’t have to manage servers, it can increase your size with the number of requests and you only pay for what you use.</p><p>The idea behind microservices is to divide your monolith application into small services, so you can develop, increase, and manage independently.<br>O FaaS (Function as a Service) divides into even small things. So, we can see a tendency that the unity of work its being smaller and we evolve from monolith to microservices and now we have functions.</p><p>The biggest difference between Chalice and the other ones is that it’s focused on the use of a decorator-based API to write applications without servers and is executed in AWS Lambda. Your goal is to make the development and deployment of these types of apps as simple as possible.</p><p>To reach this goal, Chalice needs to guess some things about how the apps will be deployed and make restrictions for the app structure. The set of resources is smaller and your documentation is very simple too.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*NOtY8qNZYa5VshSm" /><figcaption>Chalice Logo</figcaption></figure><p>An alternative to Chalice is Zappa, which is a Python Serverless Framework with a simple deployment process, and it’s very similar. We have a very strong discussion in the community comparing your main features. The decision of which one you should use depends on your project and what benefits it needs to reach.</p><p>These options aren’t perfect and both have big compensations.</p><p><strong>Choose Chalice if:</strong><br>- You need a specific feature that Chalices provides.<br>- You are starting an API from zero.<br>- You want just to make it simple</p><p><strong>Choose Zappa if:</strong><br>- You imagine that your API will be big.<br>- You already have an API in Flask or Django.<br>- Your API it’s essential for the product.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/200/0*BlXLjuzGsCe8OXhc" /><figcaption>Zappa Logo</figcaption></figure><p>The next, but not least, is the Serverless which works with AWS Lambda too. To create a project and deploy is very easy, your CLI provides all necessary commands for the process. Its structure is defined in serverless.yaml file, here is an example:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*SlhW3pswehgpO1KL.png" /></figure><p>It’s possible to define multiple runtimes in a unique serverless microservice, we can use different languages in different functions. Furthermore, we can do communication between and make a complete frontend and backend application, for example.</p><p>About the monitoring, Serverless has a default one and a beta service called Serverless Console, which can be configured with a simple flag in the .yml file, and has the objective of providing more precisely metrics.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*d-cyY_7wL0VdH82d.png" /><figcaption>Serverless Logo</figcaption></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e090bab1757e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[A comparative analysis about methods to start your model deployment lifecycle. [BentoML and KServe]]]></title>
            <link>https://medium.com/@vtmacedo/a-comparative-analysis-about-methods-to-start-your-model-deployment-lifecycle-bentoml-and-kserve-1c1517144e63?source=rss-48b407ffbdb6------2</link>
            <guid isPermaLink="false">https://medium.com/p/1c1517144e63</guid>
            <category><![CDATA[devops]]></category>
            <category><![CDATA[mlops-tool]]></category>
            <category><![CDATA[mlops]]></category>
            <category><![CDATA[mlops-platform]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Vinicius Torres de Macedo]]></dc:creator>
            <pubDate>Tue, 21 Mar 2023 23:56:14 GMT</pubDate>
            <atom:updated>2023-03-21T23:56:14.624Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*AIH006utNiJ1Gtf_.png" /></figure><p>In this post, I want to discuss a little bit about some options to start thinking in your MLops infrastructure. If you work in a startup or small company and your data team is still working with Google Sheets or Excel spreadsheets to manage your models, this post is for you.</p><p>Nowadays, with the popularity of many IA tools, most companies are looking for ways to increase the value of their products by adding machine learning features in order to stay competitive in the market and meet the evolving needs of their customers. Because of this reason, sooner or later they will run into model deployment issues.</p><p>The first alternative that we will talk about is <a href="https://www.kubeflow.org/"><strong>Kubeflow</strong></a>, an open-source platform for machine learning and MLOps on Kubernetes introduced by Google.</p><p>The different stages in a typical machine learning lifecycle are represented with different software components in Kubeflow, including model development, model training, model serving, and automated machine learning. Thus, let’s focus on one software component used for model serving, the <a href="https://www.kubeflow.org/docs/external-add-ons/kserve/kserve/"><strong>Kserve,</strong></a> and compare this with another open platform that simplifies ML model deployment and enables you to serve your models at production scale in minutes, <a href="https://docs.bentoml.org/en/latest/index.html"><strong>BentoML</strong></a>.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/791/1*pi1K9HUtDqybaEoNWChNLg.png" /><figcaption>BentoML and Kserve logomarks.</figcaption></figure><p>With a view to decide which is the best one, I tried to build POCs (Proof of Concepts) using both. In my opinion, BentoML wins because of their robust documentation with a lot of up-to-date examples. As I am starting to study the area, this was a crucial point for understanding the process as a whole.</p><p>Furthermore, BentoML provides many different deployment options for ML teams, the three most common options are:</p><ul><li>🐳 Generate <strong>container images</strong> from Bento for custom docker deployment</li><li><a href="https://github.com/bentoml/Yatai">🦄️ <strong>Yatai</strong></a>: Model Deployment at scale on Kubernetes</li><li><a href="https://github.com/bentoml/bentoctl">🚀 <strong>bentoctl</strong></a>: Fast model deployment on any cloud platform</li></ul><p>The simplest one is using a Docker image created to be DevOps friendly and ready for production. The bentoctl helps to deploy on the cloud, supporting AWS SageMaker, AWS Lambda, EC2, Google Compute Engine, Azure, Heroku and more. Despite of that, my favorite method is Yatai, which provides a good user interface to manage and lets you deploy, operate and scale services on Kubernetes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*XP3aSeXUuRoDURjU.png" /><figcaption>Yatai UI</figcaption></figure><p>On the other hand, Kserve has some advantages like compatibility with canary deployments , scale-to-zero and autoscaling. We can define Kserve as a server orchestrators, which add orchestrating features for scaling, deploying and generally managing the server fleet, that is why it is a more complex tool to use. Since, BentoML is a inference server, which deal with wrapping the model into a microservice.</p><p>My personal suggestion is to try out the quickstart tutorials for each and see what fits your needs best. Both tools are good to start thinking about simplifying the ML process, however, it’s necessary that your team already do these things manually and have a sense of the process as a whole.</p><p>Currently, with the need for rapid evolution, many companies implement these processes in a simple notebook, and it works. So, the decision to implement a tool like this is similar to building a continuous deployment pipeline, it requires work at first, but then the benefits pay off.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1c1517144e63" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>