<?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 Manish Kumar on Medium]]></title>
        <description><![CDATA[Stories by Manish Kumar on Medium]]></description>
        <link>https://medium.com/@perceptronn?source=rss-ba39330d2a50------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*heXxWkjloa6U5mK0.</url>
            <title>Stories by Manish Kumar on Medium</title>
            <link>https://medium.com/@perceptronn?source=rss-ba39330d2a50------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 10:49:38 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@perceptronn/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[Unveiling the Potential of Positive and Unlabeled Learning: A Deep Dive into PU Modeling]]></title>
            <link>https://medium.com/@perceptronn/unveiling-the-potential-of-positive-and-unlabeled-learning-a-deep-dive-into-pu-modeling-516b240b8573?source=rss-ba39330d2a50------2</link>
            <guid isPermaLink="false">https://medium.com/p/516b240b8573</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[life-sciences]]></category>
            <category><![CDATA[transactional-database]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Manish Kumar]]></dc:creator>
            <pubDate>Mon, 23 Oct 2023 13:34:52 GMT</pubDate>
            <atom:updated>2023-10-23T13:35:32.810Z</atom:updated>
            <content:encoded><![CDATA[<p>In a world overwhelmed with data, extracting meaningful insights is paramount. One such avenue is Positive and Unlabeled Learning (PU Learning), a niche yet potent machine learning paradigm. This method is particularly beneficial when labeled negative examples are scarce or absent, but there’s an abundance of positive and unlabeled examples.</p><p><strong>Mathematical Core:</strong></p><p>At the heart of PU Learning is the principle of identifying the binary classification boundary by solely leveraging positive and unlabeled data. Mathematically, the objective of PU learning can be depicted as minimizing the following risk function:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/975/1*XJxhWdlhVpUthpfi4-arkg.png" /></figure><p><strong>Step-by-Step Process of PU Learning:</strong></p><p><strong>Data Preparation:</strong></p><ul><li>Positive Examples Collection: Collect a substantial number of positive examples relevant to the task at hand.</li><li>Unlabeled Data Amassment: Gather a larger dataset of unlabeled examples.</li></ul><p><strong>Model Initialization:</strong></p><ul><li>Choose a suitable model and initialize it with random or heuristic-based parameters.</li></ul><p><strong>Training:</strong></p><ul><li>Positive Example Training: Train the model initially with positive examples.</li><li>Unlabeled Data Training: Gradually introduce unlabeled data, adjusting the model to minimize the loss function, typically by treating unlabeled examples as negative.</li></ul><p><strong>Model Evaluation and Tuning:</strong></p><ul><li>Evaluate the model’s performance using metrics like precision, recall, or F1 score.</li><li>Tune hyperparameters and adjust the loss function to improve model performance.</li></ul><p><strong>Do’s and Don’ts:</strong></p><p><strong>DOs:</strong></p><ol><li>Ensure a Considerable Amount of Positive Examples.</li><li>Utilize Reliable Unlabeled Data.</li><li>Experiment with Different Loss Functions.</li></ol><p><strong>DONTs:</strong></p><ol><li>Avoid Overfitting.</li><li>Don’t Neglect Feature Engineering.</li></ol><p><strong>Industry Applications:</strong></p><p><strong>Retail:</strong></p><p>In the retail sector, PU Learning can be instrumental in customer segmentation, identifying potential customers amidst a vast pool of unlabeled data.</p><p><strong>Pharmaceuticals:</strong></p><p>In pharmaceuticals, PU Learning can expedite drug discovery by identifying potential drug compounds from a myriad of unlabeled compounds. It also be used to identify patients at risk of certain diseases and predicting any missing event in patient journey.</p><p><strong>In conclusion</strong>, PU Learning opens up new horizons in machine learning, particularly where negative examples are hard to come by. Its mathematical elegance coupled with practical utility across sectors like retail and pharmaceuticals makes it a worthy contender in the ML toolkit. The journey of mastering PU Learning is filled with both challenges and opportunities, echoing the broader narrative of machine learning’s quest for mimicking human-like learning from sparse or incomplete information.</p><p>Reference: Jaskie, M. (2019). Applications and Approaches to the Positive and Unlabeled Learning Problem: A Survey. Retrieved from <a href="https://sensip.engineering.asu.edu/wp-content/uploads/2020/03/Jaskie_IISA_2019_Applications-and-Approaches-to-the-Positive-and-Unlabeled-Learning-problem-A-Survey-Final-Version.pdf">ASU Engineering</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=516b240b8573" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>