Online Continual Learning under Extreme Memory Constraints

TL;DR

This paper introduces Memory-Constrained Online Continual Learning (MC-OCL), a new problem setting with strict memory limits, and proposes Batch-level Distillation (BLD), a regularization-based method that balances learning new tasks and retaining old knowledge under these constraints.

Contribution

The paper defines the MC-OCL problem and proposes BLD, a novel regularization-based algorithm that operates effectively under extreme memory constraints.

Findings

BLD achieves comparable accuracy to higher-memory distillation methods.

Experimental results on three benchmarks validate the effectiveness of BLD.

MC-OCL is a viable new setting for continual learning research.

Abstract

Continual Learning (CL) aims to develop agents emulating the human ability to sequentially learn new tasks while being able to retain knowledge obtained from past experiences. In this paper, we introduce the novel problem of Memory-Constrained Online Continual Learning (MC-OCL) which imposes strict constraints on the memory overhead that a possible algorithm can use to avoid catastrophic forgetting. As most, if not all, previous CL methods violate these constraints, we propose an algorithmic solution to MC-OCL: Batch-level Distillation (BLD), a regularization-based CL approach, which effectively balances stability and plasticity in order to learn from data streams, while preserving the ability to solve old tasks through distillation. Our extensive experimental evaluation, conducted on three publicly available benchmarks, empirically demonstrates that our approach successfully addresses the MC-OCL problem and achieves comparable accuracy to prior distillation methods requiring higher memory overhead.