BarlowRL: Barlow Twins for Data-Efficient Reinforcement Learning

TL;DR

BarlowRL integrates Barlow Twins self-supervised learning with DER to improve data efficiency and performance in reinforcement learning, especially on Atari benchmarks, by promoting well-spread state representations.

Contribution

This work introduces BarlowRL, a novel reinforcement learning agent that combines Barlow Twins with DER to enhance data efficiency and avoid dimensional collapse.

Findings

Outperforms DER and CURL on Atari 100k benchmark

Achieves better data efficiency in RL tasks

Utilizes self-supervised learning to improve state representations

Abstract

This paper introduces BarlowRL, a data-efficient reinforcement learning agent that combines the Barlow Twins self-supervised learning framework with DER (Data-Efficient Rainbow) algorithm. BarlowRL outperforms both DER and its contrastive counterpart CURL on the Atari 100k benchmark. BarlowRL avoids dimensional collapse by enforcing information spread to the whole space. This helps RL algorithms to utilize uniformly spread state representation that eventually results in a remarkable performance. The integration of Barlow Twins with DER enhances data efficiency and achieves superior performance in the RL tasks. BarlowRL demonstrates the potential of incorporating self-supervised learning techniques to improve RL algorithms.