Self-Sustaining Representation Expansion for Non-Exemplar Class-Incremental Learning

TL;DR

This paper introduces a self-sustaining representation expansion method for non-exemplar class-incremental learning, effectively maintaining old features and improving recognition of old and new classes without storing old samples.

Contribution

It proposes a novel structure reorganization and distillation scheme with prototype selection to enhance class discrimination in non-exemplar incremental learning.

Findings

Outperforms state-of-the-art methods by 3-6% on three benchmarks.

Effectively maintains old features without old class samples.

Improves discrimination between old and new classes.

Abstract

Non-exemplar class-incremental learning is to recognize both the old and new classes when old class samples cannot be saved. It is a challenging task since representation optimization and feature retention can only be achieved under supervision from new classes. To address this problem, we propose a novel self-sustaining representation expansion scheme. Our scheme consists of a structure reorganization strategy that fuses main-branch expansion and side-branch updating to maintain the old features, and a main-branch distillation scheme to transfer the invariant knowledge. Furthermore, a prototype selection mechanism is proposed to enhance the discrimination between the old and new classes by selectively incorporating new samples into the distillation process. Extensive experiments on three benchmarks demonstrate significant incremental performance, outperforming the state-of-the-art methods by a margin of 3%, 3% and 6%, respectively.