La-MAML: Look-ahead Meta Learning for Continual Learning

TL;DR

La-MAML introduces a fast, online meta-learning algorithm with a small episodic memory that effectively mitigates catastrophic forgetting in continual learning, outperforming existing methods on visual classification benchmarks.

Contribution

It proposes La-MAML, a novel meta-learning algorithm with per-parameter learning rate modulation for online continual learning, improving efficiency and performance over prior approaches.

Findings

La-MAML outperforms existing continual learning methods on visual benchmarks.

The algorithm effectively reduces catastrophic forgetting.

It demonstrates fast online adaptation with a small episodic memory.

Abstract

The continual learning problem involves training models with limited capacity to perform well on a set of an unknown number of sequentially arriving tasks. While meta-learning shows great potential for reducing interference between old and new tasks, the current training procedures tend to be either slow or offline, and sensitive to many hyper-parameters. In this work, we propose Look-ahead MAML (La-MAML), a fast optimisation-based meta-learning algorithm for online-continual learning, aided by a small episodic memory. Our proposed modulation of per-parameter learning rates in our meta-learning update allows us to draw connections to prior work on hypergradients and meta-descent. This provides a more flexible and efficient way to mitigate catastrophic forgetting compared to conventional prior-based methods. La-MAML achieves performance superior to other replay-based, prior-based and meta-learning based approaches for continual learning on real-world visual classification benchmarks. Source code can be found here: https://github.com/montrealrobotics/La-MAML