<?xml version="1.0" encoding="utf-8" standalone="yes" ?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Machine Learning on Gintare Karolina Dziugaite</title>
    <link>/tags/machine-learning/</link>
    <description>Recent content in Machine Learning on Gintare Karolina Dziugaite</description>
    <generator>Hugo -- gohugo.io</generator>
    <language>en-us</language>
    <copyright>&amp;copy; 2018</copyright>
    <lastBuildDate>Sun, 01 Mar 2026 00:00:00 +0000</lastBuildDate>
    
	<atom:link href="/tags/machine-learning/index.xml" rel="self" type="application/rss+xml" />
    
    
    <item>
      <title>The Interplay between Memorization and Generalization in Deep Learning</title>
      <link>/post/memorization-vs-generalization/</link>
      <pubDate>Sun, 01 Mar 2026 00:00:00 +0000</pubDate>
      
      <guid>/post/memorization-vs-generalization/</guid>
      <description>Download PDF
Introduction: The Open Secret of Deep Learning It is an open secret in modern machine learning: neural networks memorize their training data.1 We are no longer just theorizing about this; there is striking empirical evidence showing that researchers can reconstruct training images or extract verbatim text snippets directly from trained models. Studies like Haim et al. (2023) have reconstructed recognizable images from standard vision models, while Carlini et al.</description>
    </item>
    
  </channel>
</rss>