Mitigating Forgetting in Online Continual Learning via Contrasting Semantically Distinct Augmentations

TL;DR

This paper proposes a novel contrastive learning approach with semantically distinct augmentations to mitigate catastrophic forgetting in online class-incremental continual learning, achieving superior results on standard datasets.

Contribution

It introduces semantically distinct augmentation in contrastive learning for OCL, improving model stability and reducing forgetting compared to existing methods.

Findings

Achieves state-of-the-art performance on CIFAR-10, CIFAR-100, and Mini-ImageNet.

Effectively alleviates catastrophic forgetting in class-incremental OCL.

Enhances model stability with semantically distinct augmentations.

Abstract

Online continual learning (OCL) aims to enable model learning from a non-stationary data stream to continuously acquire new knowledge as well as retain the learnt one, under the constraints of having limited system size and computational cost, in which the main challenge comes from the "catastrophic forgetting" issue -- the inability to well remember the learnt knowledge while learning the new ones. With the specific focus on the class-incremental OCL scenario, i.e. OCL for classification, the recent advance incorporates the contrastive learning technique for learning more generalised feature representation to achieve the state-of-the-art performance but is still unable to fully resolve the catastrophic forgetting. In this paper, we follow the strategy of adopting contrastive learning but further introduce the semantically distinct augmentation technique, in which it leverages strong augmentation to generate more data samples, and we show that considering these samples semantically different from their original classes (thus being related to the out-of-distribution samples) in the contrastive learning mechanism contributes to alleviate forgetting and facilitate model stability. Moreover, in addition to contrastive learning, the typical classification mechanism and objective (i.e. softmax classifier and cross-entropy loss) are included in our model design for faster convergence and utilising the label information, but particularly equipped with a sampling strategy to tackle the tendency of favouring the new classes (i.e. model bias towards the recently learnt classes). Upon conducting extensive experiments on CIFAR-10, CIFAR-100, and Mini-Imagenet datasets, our proposed method is shown to achieve superior performance against various baselines.