Slowing Down Forgetting in Continual Learning

TL;DR

This paper introduces ReCL, a novel framework that reduces catastrophic forgetting in continual learning by leveraging the implicit bias of neural networks to reconstruct old data, improving performance across various scenarios.

Contribution

ReCL is the first framework to use neural networks' convergence properties to reconstruct previous data and mitigate forgetting in continual learning.

Findings

ReCL significantly improves performance in class and domain incremental learning.

ReCL demonstrates large gains across multiple datasets and architectures.

The framework is compatible with existing continual learning methods.

Abstract

A common challenge in continual learning (CL) is catastrophic forgetting, where the performance on old tasks drops after new, additional tasks are learned. In this paper, we propose a novel framework called ReCL to slow down forgetting in CL. Our framework exploits an implicit bias of gradient-based neural networks due to which these converge to margin maximization points. Such convergence points allow us to reconstruct old data from previous tasks, which we then combine with the current training data. Our framework is flexible and can be applied on top of existing, state-of-the-art CL methods. We further demonstrate the performance gain from our framework across a large series of experiments, including two challenging CL scenarios (class incremental and domain incremental learning), different datasets (MNIST, CIFAR10, TinyImagenet), and different network architectures. Across all experiments, we find large performance gains through ReCL. To the best of our knowledge, our framework is the first to address catastrophic forgetting by leveraging models in CL as their own memory buffers.