Understanding Imbalanced Forgetting in Rehearsal-Based Class-Incremental Learning

2026-05-14

AI Summary

In replay-based class-incremental learning, imbalanced forgetting of old classes persists even when replay samples are evenly distributed. This work is the first to investigate this phenomenon through the lens of gradient interference, introducing three interference coefficients derived from the final layer of the neural network to systematically characterize the gradient-level interference experienced by each class during training. The study reveals that self-induced interference is the primary driver of forgetting, while new-class interference modulates its effect. Through controlled experiments and correlation analyses, the authors establish an interpretable link between interference sources and the degree of forgetting. The proposed coefficients reliably predict the forgetting ranking of old classes at the end of each incremental step, offering a mechanistic understanding and a principled basis for mitigating imbalanced forgetting.

Abstract

Neural networks suffer from catastrophic forgetting in class-incremental learning (CIL) settings. Rehearsal$\unicode{x2013}$replaying a subset of past samples$\unicode{x2013}$is a well-established mitigation strategy. However, recent results suggest that, despite balanced rehearsal allocation, some classes are forgotten substantially more than others. Despite its relevance, this imbalanced forgetting phenomenon remains underexplored. This work shows that imbalanced forgetting arises systematically and severely in rehearsal-based CIL and investigates it extensively. Specifically, we construct, from a principled analysis, three last-layer coefficients that capture different gradient-level sources of interference affecting each past class during an incremental step. We then demonstrate that, together, they reliably predict how past classes will rank in terms of forgetting at the end of that step. While predictive performance alone does not establish causality, these results support the interpretation of the coefficients as a plausible mechanistic account linking last-layer gradient-level interactions during training to class-level forgetting outcomes. Notably, one coefficient$\unicode{x2013}$capturing self-induced interference$\unicode{x2013}$emerges as the strongest predictor, with controlled experiments providing evidence consistent with this coefficient being influenced by the new-class interference coefficient. Overall, our findings provide valuable insights and suggest promising directions for mitigating imbalanced forgetting by reducing class-wise disparities in the identified sources of interference.

Problem

• imbalanced forgetting
• class-incremental learning
• rehearsal
• catastrophic forgetting
• interference

Innovation

• imbalanced forgetting
• class-incremental learning
• rehearsal
• gradient interference
• last-layer analysis