Unlocking the Power of Rehearsal in Continual Learning: A Theoretical Perspective

Junze Deng, Qinhang Wu, Peizhong Ju, Sen Lin, Yingbin Liang, Ness Shroff

Abstract: Rehearsal-based methods have shown superior performance in addressing catastrophic forgetting in continual learning (CL) by storing and training on a subset of past data alongside new data in current task. While such a concurrent rehearsal strategy is widely used, it remains unclear if this approach is always optimal. Inspired by human learning, where sequentially revisiting tasks helps mitigate forgetting, we explore whether sequential rehearsal can offer greater benefits for CL compared to standard concurrent rehearsal. To address this question, we conduct a theoretical analysis of rehearsal-based CL in overparameterized linear models, comparing two strategies: 1) Concurrent Rehearsal, where past and new data are trained together, and 2) Sequential Rehearsal, where new data is trained first, followed by revisiting past data sequentially. By explicitly characterizing forgetting and generalization error, we show that sequential rehearsal performs better when tasks are less similar. These insights further motivate a novel Hybrid Rehearsal method, which trains similar tasks concurrently and revisits dissimilar tasks sequentially. We characterize its forgetting and generalization performance, and our experiments with deep neural networks further confirm that the hybrid approach outperforms standard concurrent rehearsal. This work provides the first comprehensive theoretical analysis of rehearsal-based CL.

Lay Summary: In continual learning (CL), models often forget old knowledge when learning new tasks. We propose a novel sequential rehearsal strategy and analyze its performance alongside the widely used concurrent rehearsal method in overparameterized models. Our theory shows that sequential rehearsal performs better when tasks are less similar. We further propose a novel hybrid rehearsal framework based on our theoretical observation. Our experiments on real-world data validates our theory, offering a more effective solution to forgetting in CL.

Primary Area: Theory->Online Learning and Bandits

Keywords: Continual Learning, rehearsal-based methods, catastrophic forgetting

Submission Number: 14014