Generative Feature Replay with Orthogonal Weight Modification for Continual Learning

TL;DR

This paper introduces a novel continual learning method combining generative feature replay with orthogonal weight modification, significantly reducing catastrophic forgetting in class incremental learning scenarios.

Contribution

It proposes replaying penultimate layer features with a generative model and using self-supervision to improve stability, outperforming existing methods including those with real data storage.

Findings

Our method outperforms traditional generative replay and OWM alone.

It achieves better accuracy than baselines on multiple datasets.

The approach effectively mitigates catastrophic forgetting in class incremental learning.

Abstract

The ability of intelligent agents to learn and remember multiple tasks sequentially is crucial to achieving artificial general intelligence. Many continual learning (CL) methods have been proposed to overcome catastrophic forgetting which results from non i.i.d data in the sequential learning of neural networks. In this paper we focus on class incremental learning, a challenging CL scenario. For this scenario, generative replay is a promising strategy which generates and replays pseudo data for previous tasks to alleviate catastrophic forgetting. However, it is hard to train a generative model continually for relatively complex data. Based on recently proposed orthogonal weight modification (OWM) algorithm which can approximately keep previously learned feature invariant when learning new tasks, we propose to 1) replay penultimate layer feature with a generative model; 2) leverage a self-supervised auxiliary task to further enhance the stability of feature. Empirical results on several datasets show our method always achieves substantial improvement over powerful OWM while conventional generative replay always results in a negative effect. Meanwhile our method beats several strong baselines including one based on real data storage. In addition, we conduct experiments to study why our method is effective.