Memorization Through the Lens of Curvature of Loss Function Around Samples

TL;DR

This paper introduces a curvature-based metric to measure memorization in deep neural networks, revealing new failure modes and correlating well with known memorization patterns, while being scalable and effective.

Contribution

It proposes a novel curvature-based measure of memorization that captures failure modes and correlates with existing metrics, offering a scalable alternative.

Findings

High curvature samples often correspond to mislabeled or conflicting data.

The method detects duplicated images with different labels as a failure mode.

Curvature scores effectively identify memorized corrupted samples.

Abstract

Deep neural networks are over-parameterized and easily overfit the datasets they train on. In the extreme case, it has been shown that these networks can memorize a training set with fully randomized labels. We propose using the curvature of loss function around each training sample, averaged over training epochs, as a measure of memorization of the sample. We use this metric to study the generalization versus memorization properties of different samples in popular image datasets and show that it captures memorization statistics well, both qualitatively and quantitatively. We first show that the high curvature samples visually correspond to long-tailed, mislabeled, or conflicting samples, those that are most likely to be memorized. This analysis helps us find, to the best of our knowledge, a novel failure mode on the CIFAR100 and ImageNet datasets: that of duplicated images with differing labels. Quantitatively, we corroborate the validity of our scores via two methods. First, we validate our scores against an independent and comprehensively calculated baseline, by showing high cosine similarity with the memorization scores released by Feldman and Zhang (2020). Second, we inject corrupted samples which are memorized by the network, and show that these are learned with high curvature. To this end, we synthetically mislabel a random subset of the dataset. We overfit a network to it and show that sorting by curvature yields high AUROC values for identifying the corrupted samples. An added advantage of our method is that it is scalable, as it requires training only a single network as opposed to the thousands trained by the baseline, while capturing the aforementioned failure mode that the baseline fails to identify.