Re-evaluating Continual Learning with Few-Shot Adaptation

Continual learning methods aim to maximize the stability and plasticity of machine learning models that are trained on a sequence of tasks. The standard measure of stability (i.e., forgetting) is the 0-shot performance of a model on previously learned tasks, and plasticity, the performance on the most recently learned task. However, 0-shot evaluation does not fully measure a model or method's ability to retain learned information or adapt quickly to new information, as it requires perfect recall across multiple tasks. In this paper, we propose *few-shot evaluation* as a more comprehensive assessment of the stability and plasticity of a continual learning system. Through few-shot evaluation with a novel metric—*per-shot plasticity*—on task sequences for continual image classification, we show that a mixture of *foresight* via the meta-learning of a short sequence of future tasks and *hindsight* via replay on previously learned tasks leads to a model that strikes an ideal balance between stability and plasticity in a continual learning setting.