Characterizing Datapoints via Second-Split Forgetting

Pratyush Maini, Saurabh Garg, Zachary Chase Lipton, J Zico Kolter

Keywords: example hardness, memorization, generalization, forgetting, unlearning

TL;DR: We analyze the forgetting and learning dynamics of neural networks to characterize different types of hard examples as belonging to mislabeled, rare and complex categories.

Abstract: Researchers investigating example hardness have increasingly focused on the dynamics by which neural networks learn and forget examples throughout training. Popular metrics derived from these dynamics include (i) the epoch at which examples are first correctly classified; (ii) the number of times their predictions flip during training; and (iii) whether their prediction flips if they are held out. However, these metrics do not distinguish among examples that are hard for distinct reasons, such as membership in a rare subpopulation, being mislabeled, or belonging to a complex subpopulation. In this paper, we propose *second-split forgetting time* (SSFT), a complementary metric that tracks the epoch (if any) after which an original training example is forgotten as the network is fine-tuned on a randomly held out partition of the data. Across multiple benchmark datasets and modalities, we demonstrate that *mislabeled* examples are forgotten quickly, and seemingly *rare* examples are forgotten comparatively slowly. By contrast, metrics only considering the first split learning dynamics struggle to differentiate the two. At large learning rates, SSFT tends to be robust across architectures, optimizers, and random seeds. From a practical standpoint, the SSFT can (i) help to identify mislabeled samples, the removal of which improves generalization; and (ii) provide insights about failure modes. Through theoretical analysis addressing overparameterized linear models, we provide insights into how the observed phenomena may arise.

Community Implementations: [ 2 code implementations](https://www.catalyzex.com/paper/arxiv:2210.15031/code)