Learning offline: memory replay in biological and artificial reinforcement learning

TL;DR

This paper reviews the role of replay in biological and artificial reinforcement learning, highlighting its importance for memory consolidation, stability, and generalisation, and explores how insights from both fields can inform each other.

Contribution

It synthesizes recent advances in understanding replay's functions in neuroscience and AI, proposing a cross-disciplinary perspective to enhance learning and memory models.

Findings

Replay supports memory consolidation in biological systems.

Replay stabilizes learning in deep neural networks.

Cross-field insights can improve reinforcement learning methods.

Abstract

Learning to act in an environment to maximise rewards is among the brain's key functions. This process has often been conceptualised within the framework of reinforcement learning, which has also gained prominence in machine learning and artificial intelligence (AI) as a way to optimise decision-making. A common aspect of both biological and machine reinforcement learning is the reactivation of previously experienced episodes, referred to as replay. Replay is important for memory consolidation in biological neural networks, and is key to stabilising learning in deep neural networks. Here, we review recent developments concerning the functional roles of replay in the fields of neuroscience and AI. Complementary progress suggests how replay might support learning processes, including generalisation and continual learning, affording opportunities to transfer knowledge across the two fields to advance the understanding of biological and artificial learning and memory.