Fast and Data-Efficient Training of Rainbow: an Experimental Study on Atari

TL;DR

This paper presents an improved, data-efficient version of Rainbow DQN that maintains competitive performance on Atari games while drastically reducing training data and time requirements, making RL research more accessible.

Contribution

We propose modifications to Rainbow that significantly reduce data and compute needs without sacrificing performance, validated through extensive experiments and ablation studies.

Findings

Achieves median human-normalized scores close to original Rainbow

Uses 20 times less data than standard Rainbow

Requires only 7.5 hours of training on a single GPU

Abstract

Across the Arcade Learning Environment, Rainbow achieves a level of performance competitive with humans and modern RL algorithms. However, attaining this level of performance requires large amounts of data and hardware resources, making research in this area computationally expensive and use in practical applications often infeasible. This paper's contribution is threefold: We (1) propose an improved version of Rainbow, seeking to drastically reduce Rainbow's data, training time, and compute requirements while maintaining its competitive performance; (2) we empirically demonstrate the effectiveness of our approach through experiments on the Arcade Learning Environment, and (3) we conduct a number of ablation studies to investigate the effect of the individual proposed modifications. Our improved version of Rainbow reaches a median human normalized score close to classic Rainbow's, while using 20 times less data and requiring only 7.5 hours of training time on a single GPU. We also provide our full implementation including pre-trained models.