Measures of Information Reflect Memorization Patterns

TL;DR

This paper introduces information-theoretic measures of neural activation diversity to detect and differentiate between heuristic and example memorization in neural networks, aiding in understanding and improving model generalization.

Contribution

It proposes a novel approach using activation diversity measures to identify memorization patterns, applicable even on unlabeled data, and demonstrates its utility for model selection.

Findings

Activation diversity correlates with generalization and memorization.

Information measures distinguish between heuristic and example memorization.

Method aids in model selection based on memorization patterns.

Abstract

Neural networks are known to exploit spurious artifacts (or shortcuts) that co-occur with a target label, exhibiting heuristic memorization. On the other hand, networks have been shown to memorize training examples, resulting in example-level memorization. These kinds of memorization impede generalization of networks beyond their training distributions. Detecting such memorization could be challenging, often requiring researchers to curate tailored test sets. In this work, we hypothesize -- and subsequently show -- that the diversity in the activation patterns of different neurons is reflective of model generalization and memorization. We quantify the diversity in the neural activations through information-theoretic measures and find support for our hypothesis on experiments spanning several natural language and vision tasks. Importantly, we discover that information organization points to the two forms of memorization, even for neural activations computed on unlabelled in-distribution examples. Lastly, we demonstrate the utility of our findings for the problem of model selection. The associated code and other resources for this work are available at https://rachitbansal.github.io/information-measures.