Three scenarios for continual learning

TLDR

Catastrophic forgetting hampers continual learning in standard neural networks, and the diversity of proposed methods and inconsistent evaluation protocols make direct comparison difficult. This work introduces three continual learning scenarios that differ by whether task identity is provided at test time and, if not, whether it must be inferred. The authors evaluate recent continual learning methods on split and permuted MNIST tasks under each scenario, performing extensive comparisons across well‑defined task sequences. They find substantial differences in difficulty and method efficiency across scenarios, and that when task identity must be inferred, regularization approaches fail while replaying past representations is necessary.

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.