Three scenarios for continual learning

TL;DR

This paper introduces three structured scenarios for evaluating continual learning methods, highlighting differences in difficulty and efficiency, especially emphasizing the challenges of class incremental learning without task identity.

Contribution

It defines three distinct continual learning scenarios based on task identity availability and inference, enabling more consistent and comparative evaluation of methods.

Findings

Regularization methods fail in class incremental learning

Replay strategies are essential for scenarios without task identity

Significant differences in method performance across scenarios

Abstract

Standard artificial neural networks suffer from the well-known issue of catastrophic forgetting, making continual or lifelong learning difficult for machine learning. In recent years, numerous methods have been proposed for continual learning, but due to differences in evaluation protocols it is difficult to directly compare their performance. To enable more structured comparisons, we describe three continual learning scenarios based on whether at test time task identity is provided and--in case it is not--whether it must be inferred. Any sequence of well-defined tasks can be performed according to each scenario. Using the split and permuted MNIST task protocols, for each scenario we carry out an extensive comparison of recently proposed continual learning methods. We demonstrate substantial differences between the three scenarios in terms of difficulty and in terms of how efficient different methods are. In particular, when task identity must be inferred (i.e., class incremental learning), we find that regularization-based approaches (e.g., elastic weight consolidation) fail and that replaying representations of previous experiences seems required for solving this scenario.