On the Stability-Plasticity Dilemma of Class-Incremental Learning

TL;DR

This paper analyzes the stability-plasticity trade-off in class-incremental learning, revealing that most algorithms favor stability, and proposes simple methods to improve feature representation learning for better balance.

Contribution

It introduces analytical tools to measure stability and plasticity, and demonstrates that current methods overly favor stability, suggesting a need for better feature learning strategies.

Findings

Most algorithms favor stability over plasticity

Feature extractors remain effective after initial training

Simple algorithms can improve feature representation analysis

Abstract

A primary goal of class-incremental learning is to strike a balance between stability and plasticity, where models should be both stable enough to retain knowledge learned from previously seen classes, and plastic enough to learn concepts from new classes. While previous works demonstrate strong performance on class-incremental benchmarks, it is not clear whether their success comes from the models being stable, plastic, or a mixture of both. This paper aims to shed light on how effectively recent class-incremental learning algorithms address the stability-plasticity trade-off. We establish analytical tools that measure the stability and plasticity of feature representations, and employ such tools to investigate models trained with various algorithms on large-scale class-incremental benchmarks. Surprisingly, we find that the majority of class-incremental learning algorithms heavily favor stability over plasticity, to the extent that the feature extractor of a model trained on the initial set of classes is no less effective than that of the final incremental model. Our observations not only inspire two simple algorithms that highlight the importance of feature representation analysis, but also suggest that class-incremental learning approaches, in general, should strive for better feature representation learning.