GradMix: Gradient-based Selective Mixup for Robust Data Augmentation in Class-Incremental Learning

TL;DR

GradMix is a novel gradient-based selective mixup data augmentation technique designed to reduce catastrophic forgetting in class-incremental learning, outperforming existing methods by intelligently mixing helpful class pairs.

Contribution

It introduces a class-based criterion for selective sample mixing, effectively mitigating knowledge loss during continual learning.

Findings

GradMix outperforms baseline data augmentation methods in accuracy.

It reduces catastrophic forgetting in class-incremental learning.

The method is validated on various real datasets.

Abstract

In the context of continual learning, acquiring new knowledge while maintaining previous knowledge presents a significant challenge. Existing methods often use experience replay techniques that store a small portion of previous task data for training. In experience replay approaches, data augmentation has emerged as a promising strategy to further improve the model performance by mixing limited previous task data with sufficient current task data. However, we theoretically and empirically analyze that training with mixed samples from random sample pairs may harm the knowledge of previous tasks and cause greater catastrophic forgetting. We then propose GradMix, a robust data augmentation method specifically designed for mitigating catastrophic forgetting in class-incremental learning. GradMix performs gradient-based selective mixup using a class-based criterion that mixes only samples from helpful class pairs and not from detrimental class pairs for reducing catastrophic forgetting. Our experiments on various real datasets show that GradMix outperforms data augmentation baselines in accuracy by minimizing the forgetting of previous knowledge.