On the Stability-Plasticity Dilemma of Class-Incremental Learning
This work addresses the stability-plasticity trade-off in class-incremental learning, which is crucial for developing more effective lifelong learning systems, though it is incremental in nature.
The paper investigates how class-incremental learning algorithms balance stability and plasticity, finding that most heavily favor stability, with the initial feature extractor often performing as well as the final incremental model.
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.