Atari-5: Distilling the Arcade Learning Environment down to Five Games

Matthew Aitchison, Penny Sweetser, Marcus Hutter
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:421-438, 2023.

Abstract

The Arcade Learning Environment (ALE) has become an essential benchmark for assessing the performance of reinforcement learning algorithms. However, the computational cost of generating results on the entire 57-game dataset limits ALE’s use and makes the reproducibility of many results infeasible. We propose a novel solution to this problem in the form of a principled methodology for selecting small but representative subsets of environments within a benchmark suite. We applied our method to identify a subset of five ALE games, we call Atari-5, which produces 57-game median score estimates within 10% of their true values. Extending the subset to 10-games recovers 80% of the variance for log-scores for all games within the 57-game set. We show this level of compression is possible due to a high degree of correlation between many of the games in ALE.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-aitchison23a, title = {{A}tari-5: Distilling the Arcade Learning Environment down to Five Games}, author = {Aitchison, Matthew and Sweetser, Penny and Hutter, Marcus}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {421--438}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/aitchison23a/aitchison23a.pdf}, url = {https://proceedings.mlr.press/v202/aitchison23a.html}, abstract = {The Arcade Learning Environment (ALE) has become an essential benchmark for assessing the performance of reinforcement learning algorithms. However, the computational cost of generating results on the entire 57-game dataset limits ALE’s use and makes the reproducibility of many results infeasible. We propose a novel solution to this problem in the form of a principled methodology for selecting small but representative subsets of environments within a benchmark suite. We applied our method to identify a subset of five ALE games, we call Atari-5, which produces 57-game median score estimates within 10% of their true values. Extending the subset to 10-games recovers 80% of the variance for log-scores for all games within the 57-game set. We show this level of compression is possible due to a high degree of correlation between many of the games in ALE.} }
Endnote
%0 Conference Paper %T Atari-5: Distilling the Arcade Learning Environment down to Five Games %A Matthew Aitchison %A Penny Sweetser %A Marcus Hutter %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-aitchison23a %I PMLR %P 421--438 %U https://proceedings.mlr.press/v202/aitchison23a.html %V 202 %X The Arcade Learning Environment (ALE) has become an essential benchmark for assessing the performance of reinforcement learning algorithms. However, the computational cost of generating results on the entire 57-game dataset limits ALE’s use and makes the reproducibility of many results infeasible. We propose a novel solution to this problem in the form of a principled methodology for selecting small but representative subsets of environments within a benchmark suite. We applied our method to identify a subset of five ALE games, we call Atari-5, which produces 57-game median score estimates within 10% of their true values. Extending the subset to 10-games recovers 80% of the variance for log-scores for all games within the 57-game set. We show this level of compression is possible due to a high degree of correlation between many of the games in ALE.
APA
Aitchison, M., Sweetser, P. & Hutter, M.. (2023). Atari-5: Distilling the Arcade Learning Environment down to Five Games. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:421-438 Available from https://proceedings.mlr.press/v202/aitchison23a.html.

Related Material