Online Continual Learning Via Candidates Voting

TL;DR

This paper proposes a memory-efficient online continual learning method for class-incremental image classification, using candidate selection and stored feature embeddings, achieving superior results with less memory on benchmarks like CIFAR-10, CIFAR-100, and CORE-50.

Contribution

It introduces a novel online continual learning approach that avoids storing original data by using feature embeddings and candidate selection, improving performance and memory efficiency.

Findings

Achieves state-of-the-art results on CIFAR-10, CIFAR-100, and CORE-50.

Requires significantly less memory than existing methods.

Effective in class-incremental learning scenarios.

Abstract

Continual learning in online scenario aims to learn a sequence of new tasks from data stream using each data only once for training, which is more realistic than in offline mode assuming data from new task are all available. However, this problem is still under-explored for the challenging class-incremental setting in which the model classifies all classes seen so far during inference. Particularly, performance struggles with increased number of tasks or additional classes to learn for each task. In addition, most existing methods require storing original data as exemplars for knowledge replay, which may not be feasible for certain applications with limited memory budget or privacy concerns. In this work, we introduce an effective and memory-efficient method for online continual learning under class-incremental setting through candidates selection from each learned task together with prior incorporation using stored feature embeddings instead of original data as exemplars. Our proposed method implemented for image classification task achieves the best results under different benchmark datasets for online continual learning including CIFAR-10, CIFAR-100 and CORE-50 while requiring much less memory resource compared with existing works.