Exploiting generalisation symmetries in accuracy-based learning classifier systems: An initial study

TL;DR

This paper explores how incorporating multi-action rules in learning classifier systems can leverage problem symmetries to enhance reinforcement learning performance without sacrificing accuracy.

Contribution

It introduces a novel approach of using multi-action rules to exploit symmetries in the state-action space within accuracy-based classifier systems.

Findings

Exploiting symmetries improves learning performance.

Multi-action rules maintain accuracy across actions.

Performance remains stable when symmetries are reduced.

Abstract

Modern learning classifier systems typically exploit a niched genetic algorithm to facilitate rule discovery. When used for reinforcement learning, such rules represent generalisations over the state-action-reward space. Whilst encouraging maximal generality, the niching can potentially hinder the formation of generalisations in the state space which are symmetrical, or very similar, over different actions. This paper introduces the use of rules which contain multiple actions, maintaining accuracy and reward metrics for each action. It is shown that problem symmetries can be exploited, improving performance, whilst not degrading performance when symmetries are reduced.