On the importance of cross-task features for class-incremental learning

TL;DR

This paper investigates the role of cross-task features in class-incremental learning, showing that improving feature quality and transfer between tasks is crucial for better performance, especially with limited data.

Contribution

It introduces an analysis of cross-task feature learning, proposes a new forgetting measure, and emphasizes the importance of feature quality over mere forgetting prevention.

Findings

Forgetting is not the main cause of low performance.

Enhancing cross-task features improves knowledge transfer.

Future methods should focus on feature quality and transfer, not just forgetting.

Abstract

In class-incremental learning, an agent with limited resources needs to learn a sequence of classification tasks, forming an ever growing classification problem, with the constraint of not being able to access data from previous tasks. The main difference with task-incremental learning, where a task-ID is available at inference time, is that the learner also needs to perform cross-task discrimination, i.e. distinguish between classes that have not been seen together. Approaches to tackle this problem are numerous and mostly make use of an external memory (buffer) of non-negligible size. In this paper, we ablate the learning of cross-task features and study its influence on the performance of basic replay strategies used for class-IL. We also define a new forgetting measure for class-incremental learning, and see that forgetting is not the principal cause of low performance. Our experimental results show that future algorithms for class-incremental learning should not only prevent forgetting, but also aim to improve the quality of the cross-task features, and the knowledge transfer between tasks. This is especially important when tasks contain limited amount of data.