Learning to Learn without Forgetting using Attention

TL;DR

This paper introduces a meta-learned, attention-based transformer optimizer designed to improve continual learning by selectively updating model parameters, thereby reducing forgetting and enhancing knowledge transfer across tasks.

Contribution

It presents a novel transformer-based meta-learning approach that learns effective weight updates to prevent catastrophic forgetting in continual learning scenarios.

Findings

Effective on benchmark datasets like SplitMNIST and SplitCIFAR-100

Improves both forward and backward transfer in continual learning

Performs well even with small labeled datasets

Abstract

Continual learning (CL) refers to the ability to continually learn over time by accommodating new knowledge while retaining previously learned experience. While this concept is inherent in human learning, current machine learning methods are highly prone to overwrite previously learned patterns and thus forget past experience. Instead, model parameters should be updated selectively and carefully, avoiding unnecessary forgetting while optimally leveraging previously learned patterns to accelerate future learning. Since hand-crafting effective update mechanisms is difficult, we propose meta-learning a transformer-based optimizer to enhance CL. This meta-learned optimizer uses attention to learn the complex relationships between model parameters across a stream of tasks, and is designed to generate effective weight updates for the current task while preventing catastrophic forgetting on previously encountered tasks. Evaluations on benchmark datasets like SplitMNIST, RotatedMNIST, and SplitCIFAR-100 affirm the efficacy of the proposed approach in terms of both forward and backward transfer, even on small sets of labeled data, highlighting the advantages of integrating a meta-learned optimizer within the continual learning framework.