Rehearsal revealed: The limits and merits of revisiting samples in continual learning

TL;DR

This paper investigates the strengths and limitations of rehearsal in continual learning, revealing how it can lead to overfitting on sample memory and affect generalization, supported by empirical evidence across benchmarks.

Contribution

It provides novel insights into the dynamics of rehearsal in continual learning, highlighting its potential to cause overfitting and its impact on model generalization.

Findings

Models tend to stay in low-loss regions after training with rehearsal.

Rehearsal can cause overfitting on sample memory, harming generalization.

Empirical evidence from three benchmarks supports these behaviors.

Abstract

Learning from non-stationary data streams and overcoming catastrophic forgetting still poses a serious challenge for machine learning research. Rather than aiming to improve state-of-the-art, in this work we provide insight into the limits and merits of rehearsal, one of continual learning's most established methods. We hypothesize that models trained sequentially with rehearsal tend to stay in the same low-loss region after a task has finished, but are at risk of overfitting on its sample memory, hence harming generalization. We provide both conceptual and strong empirical evidence on three benchmarks for both behaviors, bringing novel insights into the dynamics of rehearsal and continual learning in general. Finally, we interpret important continual learning works in the light of our findings, allowing for a deeper understanding of their successes.