<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Data-Efficient Image Recognition with Contrastive Predictive Coding · HackerPapers discussions</title><link>https://hackerpaper.kaipfister.de/p/1905.09272/</link><description>Collected HN comments and submissions linking this paper.</description><language>en</language><lastBuildDate>Mon, 02 Nov 2026 11:16:22 GMT</lastBuildDate><atom:link href="https://hackerpaper.kaipfister.de/p/1905.09272/feed.xml" rel="self" type="application/rss+xml" /><item><title>Data-Efficient Image Recognition with Contrastive Predictive Coding — mentioned by ankeshanand</title><link>https://news.ycombinator.com/item?id=22215554</link><guid isPermaLink="false">urn:hackerpapers:mention:1905.09272:22215554</guid><pubDate>Sun, 02 Feb 2020 06:47:34 GMT</pubDate><description>&lt;p&gt;Sorry if it wasn&amp;#x27;t clear, I do mention the linear classification protocol several times in the post. If you want to evaluate performance on a classification task, you have to show it labels during evaluation, otherwise it&amp;#x27;s an impossible task. Note that the encoder is freezed during evaluation, and only a linear classifier is trained on top. Now, even when evaluated on a limited set of labels (as low as 1%), contrastive pretraining outperforms purely supervised training by a large margin (check out Figure 1 in the Data-Efficient CPC paper: https://arxiv.org/abs/1905.09272. I did not get the…&lt;/p&gt;&lt;p&gt;&lt;a href="https://hackerpaper.kaipfister.de/p/1905.09272/"&gt;Paper and discussion history&lt;/a&gt;. A mention is not an endorsement.&lt;/p&gt;</description><category>cs.CV</category></item><item><title>Data-Efficient Image Recognition with Contrastive Predictive Coding — HN submission</title><link>https://news.ycombinator.com/item?id=21751244</link><guid isPermaLink="false">urn:hackerpapers:submission:1905.09272:21751244</guid><pubDate>Tue, 10 Dec 2019 10:20:46 GMT</pubDate><description>&lt;p&gt;Data-Efficient Image Recognition with Contrastive Predictive Coding&lt;/p&gt;&lt;p&gt;&lt;/p&gt;&lt;p&gt;2 points; 1 discussion reply on HN. &lt;a href="https://hackerpaper.kaipfister.de/p/1905.09272/"&gt;Paper and source history&lt;/a&gt;. A submission is not an endorsement.&lt;/p&gt;</description><category>cs.CV</category></item><item><title>Data-Efficient Image Recognition with Contrastive Predictive Coding — HN submission</title><link>https://news.ycombinator.com/item?id=21740473</link><guid isPermaLink="false">urn:hackerpapers:submission:1905.09272:21740473</guid><pubDate>Mon, 09 Dec 2019 04:51:37 GMT</pubDate><description>&lt;p&gt;Big gains on data-efficiency and transfer learning on images (self-supervision)&lt;/p&gt;&lt;p&gt;&lt;/p&gt;&lt;p&gt;2 points; 0 discussion replies on HN. &lt;a href="https://hackerpaper.kaipfister.de/p/1905.09272/"&gt;Paper and source history&lt;/a&gt;. A submission is not an endorsement.&lt;/p&gt;</description><category>cs.CV</category></item></channel></rss>
