Deep Reinforcement Learning with Swin Transformers

TL;DR

This paper introduces Swin DQN, an online reinforcement learning method using Swin Transformers, demonstrating superior performance across 49 Atari games compared to baseline methods.

Contribution

It presents the first online RL scheme based on Swin Transformers, achieving significantly higher scores on most tested games.

Findings

Achieves higher maximal scores than baseline in 92% of games.

Attains higher mean scores than baseline in 82% of games.

Demonstrates the effectiveness of Swin Transformers in reinforcement learning.

Abstract

Transformers are neural network models that utilize multiple layers of self-attention heads and have exhibited enormous potential in natural language processing tasks. Meanwhile, there have been efforts to adapt transformers to visual tasks of machine learning, including Vision Transformers and Swin Transformers. Although some researchers use Vision Transformers for reinforcement learning tasks, their experiments remain at a small scale due to the high computational cost. This article presents the first online reinforcement learning scheme that is based on Swin Transformers: Swin DQN. In contrast to existing research, our novel approach demonstrate the superior performance with experiments on 49 games in the Arcade Learning Environment. The results show that our approach achieves significantly higher maximal evaluation scores than the baseline method in 45 of all the 49 games (92%), and higher mean evaluation scores than the baseline method in 40 of all the 49 games (82%).