Foolproof Cooperative Learning

Alexis Jacq, Julien Perolat, Matthieu Geist, Olivier Pietquin
Proceedings of The 12th Asian Conference on Machine Learning, PMLR 129:401-416, 2020.

Abstract

This paper extends the notion of learning algorithms and learning equilibriums from repeated games theory to stochastic games. We introduce Foolproof Cooperative Learning (FCL), an algorithm that converges to an equilibrium strategy that allows cooperative strategies in self-play setting while being not exploitable by selfish learners. By construction, FCL is a learning equilibrium for repeated symmetric games. We illustrate the behavior of FCL on symmetric matrix and grid games, and its robustness to selfish learners.

Cite this Paper

BibTeX
@InProceedings{pmlr-v129-jacq20a, title = {Foolproof Cooperative Learning}, author = {Jacq, Alexis and Perolat, Julien and Geist, Matthieu and Pietquin, Olivier}, booktitle = {Proceedings of The 12th Asian Conference on Machine Learning}, pages = {401--416}, year = {2020}, editor = {Pan, Sinno Jialin and Sugiyama, Masashi}, volume = {129}, series = {Proceedings of Machine Learning Research}, month = {18--20 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v129/jacq20a/jacq20a.pdf}, url = {https://proceedings.mlr.press/v129/jacq20a.html}, abstract = {This paper extends the notion of learning algorithms and learning equilibriums from repeated games theory to stochastic games. We introduce Foolproof Cooperative Learning (FCL), an algorithm that converges to an equilibrium strategy that allows cooperative strategies in self-play setting while being not exploitable by selfish learners. By construction, FCL is a learning equilibrium for repeated symmetric games. We illustrate the behavior of FCL on symmetric matrix and grid games, and its robustness to selfish learners. } }
Endnote
%0 Conference Paper %T Foolproof Cooperative Learning %A Alexis Jacq %A Julien Perolat %A Matthieu Geist %A Olivier Pietquin %B Proceedings of The 12th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Sinno Jialin Pan %E Masashi Sugiyama %F pmlr-v129-jacq20a %I PMLR %P 401--416 %U https://proceedings.mlr.press/v129/jacq20a.html %V 129 %X This paper extends the notion of learning algorithms and learning equilibriums from repeated games theory to stochastic games. We introduce Foolproof Cooperative Learning (FCL), an algorithm that converges to an equilibrium strategy that allows cooperative strategies in self-play setting while being not exploitable by selfish learners. By construction, FCL is a learning equilibrium for repeated symmetric games. We illustrate the behavior of FCL on symmetric matrix and grid games, and its robustness to selfish learners.
APA
Jacq, A., Perolat, J., Geist, M. & Pietquin, O.. (2020). Foolproof Cooperative Learning. Proceedings of The 12th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 129:401-416 Available from https://proceedings.mlr.press/v129/jacq20a.html.

Related Material