A Closer Look at Memorization in Deep Networks

TL;DR

This paper investigates how deep neural networks memorize data, showing they prioritize simple patterns, and explores how regularization affects memorization of noise versus real data, challenging traditional capacity-based explanations.

Contribution

It reveals that dataset-dependent factors influence memorization and generalization, and demonstrates how regularization can selectively degrade noise memorization without harming real data performance.

Findings

Deep networks tend to learn simple patterns before memorizing noise.

Regularization like dropout can reduce noise memorization without affecting real data accuracy.

Effective capacity notions are insufficient to explain generalization in gradient-trained deep networks.

Abstract

We examine the role of memorization in deep learning, drawing connections to capacity, generalization, and adversarial robustness. While deep networks are capable of memorizing noise data, our results suggest that they tend to prioritize learning simple patterns first. In our experiments, we expose qualitative differences in gradient-based optimization of deep neural networks (DNNs) on noise vs. real data. We also demonstrate that for appropriately tuned explicit regularization (e.g., dropout) we can degrade DNN training performance on noise datasets without compromising generalization on real data. Our analysis suggests that the notions of effective capacity which are dataset independent are unlikely to explain the generalization performance of deep networks when trained with gradient based methods because training data itself plays an important role in determining the degree of memorization.