Distributionally robust games: F-divergence and learning

Dario Bauso, Jian Gao, Tembine Hamidou

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In this paper we introduce the novel framework of distributionally robust games. These are multi-player games where each player models the state of nature using a worst-case distribution, also called adversarial distribution. Thus each player's payoff depends on the other players' decisions and on the decision of a virtual player (nature) who selects an adversarial distribution of scenarios. This paper provides three main contributions. Firstly, the distributionally robust game is formulated using the statistical notions of f-divergence between two distributions, here represented by the adversarial distribution, and the exact distribution. Secondly, the complexity of the problem is significantly reduced by means of triality theory. Thirdly, stochastic Bregman learning algorithms are proposed to speedup the computation of robust equilibria. Finally, the theoretical findings are illustrated in a convex setting and its limitations are tested with a non-convex non-concave function.

Original languageEnglish (US)
Title of host publicationProceedings of the 11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017
PublisherAssociation for Computing Machinery
Pages148-155
Number of pages8
ISBN (Print)9781450363464
DOIs
StatePublished - Dec 5 2017
Event11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017 - Venice, Italy
Duration: Dec 5 2017Dec 7 2017

Other

Other11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017
CountryItaly
CityVenice
Period12/5/1712/7/17

Fingerprint

Learning algorithms

Keywords

  • Game Theory

ASJC Scopus subject areas

  • Human-Computer Interaction
  • Computer Networks and Communications
  • Computer Vision and Pattern Recognition
  • Software

Cite this

Bauso, D., Gao, J., & Hamidou, T. (2017). Distributionally robust games: F-divergence and learning. In Proceedings of the 11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017 (pp. 148-155). Association for Computing Machinery. https://doi.org/10.1145/3150928.3150950

Distributionally robust games : F-divergence and learning. / Bauso, Dario; Gao, Jian; Hamidou, Tembine.

Proceedings of the 11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017. Association for Computing Machinery, 2017. p. 148-155.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Bauso, D, Gao, J & Hamidou, T 2017, Distributionally robust games: F-divergence and learning. in Proceedings of the 11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017. Association for Computing Machinery, pp. 148-155, 11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017, Venice, Italy, 12/5/17. https://doi.org/10.1145/3150928.3150950
Bauso D, Gao J, Hamidou T. Distributionally robust games: F-divergence and learning. In Proceedings of the 11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017. Association for Computing Machinery. 2017. p. 148-155 https://doi.org/10.1145/3150928.3150950
Bauso, Dario ; Gao, Jian ; Hamidou, Tembine. / Distributionally robust games : F-divergence and learning. Proceedings of the 11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017. Association for Computing Machinery, 2017. pp. 148-155
@inproceedings{16255cefea2c46a7bbc958b68e98bd53,
title = "Distributionally robust games: F-divergence and learning",
abstract = "In this paper we introduce the novel framework of distributionally robust games. These are multi-player games where each player models the state of nature using a worst-case distribution, also called adversarial distribution. Thus each player's payoff depends on the other players' decisions and on the decision of a virtual player (nature) who selects an adversarial distribution of scenarios. This paper provides three main contributions. Firstly, the distributionally robust game is formulated using the statistical notions of f-divergence between two distributions, here represented by the adversarial distribution, and the exact distribution. Secondly, the complexity of the problem is significantly reduced by means of triality theory. Thirdly, stochastic Bregman learning algorithms are proposed to speedup the computation of robust equilibria. Finally, the theoretical findings are illustrated in a convex setting and its limitations are tested with a non-convex non-concave function.",
keywords = "Game Theory",
author = "Dario Bauso and Jian Gao and Tembine Hamidou",
year = "2017",
month = "12",
day = "5",
doi = "10.1145/3150928.3150950",
language = "English (US)",
isbn = "9781450363464",
pages = "148--155",
booktitle = "Proceedings of the 11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017",
publisher = "Association for Computing Machinery",

}

TY - GEN

T1 - Distributionally robust games

T2 - F-divergence and learning

AU - Bauso, Dario

AU - Gao, Jian

AU - Hamidou, Tembine

PY - 2017/12/5

Y1 - 2017/12/5

N2 - In this paper we introduce the novel framework of distributionally robust games. These are multi-player games where each player models the state of nature using a worst-case distribution, also called adversarial distribution. Thus each player's payoff depends on the other players' decisions and on the decision of a virtual player (nature) who selects an adversarial distribution of scenarios. This paper provides three main contributions. Firstly, the distributionally robust game is formulated using the statistical notions of f-divergence between two distributions, here represented by the adversarial distribution, and the exact distribution. Secondly, the complexity of the problem is significantly reduced by means of triality theory. Thirdly, stochastic Bregman learning algorithms are proposed to speedup the computation of robust equilibria. Finally, the theoretical findings are illustrated in a convex setting and its limitations are tested with a non-convex non-concave function.

AB - In this paper we introduce the novel framework of distributionally robust games. These are multi-player games where each player models the state of nature using a worst-case distribution, also called adversarial distribution. Thus each player's payoff depends on the other players' decisions and on the decision of a virtual player (nature) who selects an adversarial distribution of scenarios. This paper provides three main contributions. Firstly, the distributionally robust game is formulated using the statistical notions of f-divergence between two distributions, here represented by the adversarial distribution, and the exact distribution. Secondly, the complexity of the problem is significantly reduced by means of triality theory. Thirdly, stochastic Bregman learning algorithms are proposed to speedup the computation of robust equilibria. Finally, the theoretical findings are illustrated in a convex setting and its limitations are tested with a non-convex non-concave function.

KW - Game Theory

UR - http://www.scopus.com/inward/record.url?scp=85051639774&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85051639774&partnerID=8YFLogxK

U2 - 10.1145/3150928.3150950

DO - 10.1145/3150928.3150950

M3 - Conference contribution

AN - SCOPUS:85051639774

SN - 9781450363464

SP - 148

EP - 155

BT - Proceedings of the 11th EAI International Conference on Performance Evaluation Methodologies and Tools, VALUETOOLS 2017

PB - Association for Computing Machinery

ER -