## Optimal efficient learning equilibrium: Imperfect monitoring in symmetric games (2005)

### Cached

### Download Links

- [iew3.technion.ac.il]
- [www.cs.bgu.ac.il]
- [ie.technion.ac.il]
- [www.damas.ift.ulaval.ca]
- [damas.ift.ulaval.ca]
- DBLP

### Other Repositories/Bibliography

Venue: | In Proceedings of the National Conference on Artificial Intelligence (AAAI |

Citations: | 9 - 4 self |

### BibTeX

@INPROCEEDINGS{Brafman05optimalefficient,

author = {Ronen I. Brafman and Moshe Tennenholtz},

title = {Optimal efficient learning equilibrium: Imperfect monitoring in symmetric games},

booktitle = {In Proceedings of the National Conference on Artificial Intelligence (AAAI},

year = {2005},

pages = {726--731}

}

### OpenURL

### Abstract

Efficient Learning Equilibrium (ELE) is a natural solution concept for multi-agent encounters with incomplete information. It requires the learning algorithms themselves to be in equilibrium for any game selected from a set of (initially unknown) games. In an optimal ELE, the learning algorithms would efficiently obtain the surplus the agents would obtain in an optimal Nash equilibrium of the initially unknown game which is played. The crucial part is that in an ELE deviations from the learning algorithms would become non-beneficial after polynomial time, although the game played is initially unknown. While appealing conceptually, the main challenge for establishing learning algorithms based on this concept is to isolate general classes of games where an ELE exists. Unfortunately, it has been shown that while an ELE exists for the setting in which each agent can observe all other agents ’ actions and payoffs, an ELE does not exist in general when the other agents ’ payoffs cannot be observed. In this paper we provide the first positive results on this problem, constructively proving the existence of an optimal ELE for the class of symmetric games where an agent can not observe other agents ’ payoffs. 1.

### Citations

2295 |
Game Theory
- Fudenberg, D, et al.
- 1991
(Show Context)
Citation Context ...Game-theory provides a mathematical formulation of multiagent interactions and multi-agent decision making. Here we review some of the basic concepts. For a good introduction to the area, see, e.g., (=-=Fudenberg & Tirole 1991-=-). A game in strategic form consists of a set of players I, a set of actions Ai for each i ∈ I, and a payoff function Ri : ×i∈IAi → R for each i ∈ I. We let A denote the set ×i∈IAi of joint actions. A... |

1303 | Reinforcement Learning: A Survey - Kaelbling, Littman, et al. - 1996 |

779 |
The theory of Learning in Games
- Fudenburg, Levine
- 1998
(Show Context)
Citation Context ... of researchers in cognitive psychology, experimental economics, machine learning, artificial intelligence, and related fields for quite some time (Kaelbling, Littman, & Moore 1996; Erev & Roth 1998; =-=Fudenberg & Levine 1998-=-). Much of this work uses repeated games (e.g. (Claus & Boutilier 1997; Kalai & Lehrer 1993; Conitzer & Sandholm 2003)) and stochastic games (e.g. (Littman 1994; Hu & Wellman 1998; Brafman & Tennenhol... |

498 | Markov games as a framework for multi-agent reinforcement learning
- Littman
- 1994
(Show Context)
Citation Context ...oore 1996; Erev & Roth 1998; Fudenberg & Levine 1998). Much of this work uses repeated games (e.g. (Claus & Boutilier 1997; Kalai & Lehrer 1993; Conitzer & Sandholm 2003)) and stochastic games (e.g. (=-=Littman 1994-=-; Hu & Wellman 1998; Brafman & Tennenholtz 2002; Bowling & Veloso 2001; Greenwald, Hall, & Serrano 2002)) as models of such interactions. The literature on learning in games in game theory (Fudenberg ... |

412 | Predicting How People Play Games: Reinforcement Learning in Experimental Games with Unique Mixed Strategy Equilibria
- Erev, Roth
- 1998
(Show Context)
Citation Context ...cted the attention of researchers in cognitive psychology, experimental economics, machine learning, artificial intelligence, and related fields for quite some time (Kaelbling, Littman, & Moore 1996; =-=Erev & Roth 1998-=-; Fudenberg & Levine 1998). Much of this work uses repeated games (e.g. (Claus & Boutilier 1997; Kalai & Lehrer 1993; Conitzer & Sandholm 2003)) and stochastic games (e.g. (Littman 1994; Hu & Wellman ... |

366 |
Subjectivity and correlation in randomized strategies
- Aumann
- 1974
(Show Context)
Citation Context ...ternatively, one can assume that there exists a mediator who can allow agents to correlate their strategies, by sending them private signals, as in the classical definition of correlated equilibrium (=-=Aumann 1974-=-). Although details are omitted due to lack of space, we will mention that our setting and results are easily extended to that setting of correlated equilibrium, and it allows us to extend the discuss... |

304 | The dynamics of reinforcement learning in cooperative multiagent systems
- Claus, Boutilier
- 1998
(Show Context)
Citation Context ...e learning, artificial intelligence, and related fields for quite some time (Kaelbling, Littman, & Moore 1996; Erev & Roth 1998; Fudenberg & Levine 1998). Much of this work uses repeated games (e.g. (=-=Claus & Boutilier 1997-=-; Kalai & Lehrer 1993; Conitzer & Sandholm 2003)) and stochastic games (e.g. (Littman 1994; Hu & Wellman 1998; Brafman & Tennenholtz 2002; Bowling & Veloso 2001; Greenwald, Hall, & Serrano 2002)) as m... |

283 | M.P.: Multiagent reinforcement learning: Theoretical framework and an algorithm
- Hu, Wellman
- 1998
(Show Context)
Citation Context ...v & Roth 1998; Fudenberg & Levine 1998). Much of this work uses repeated games (e.g. (Claus & Boutilier 1997; Kalai & Lehrer 1993; Conitzer & Sandholm 2003)) and stochastic games (e.g. (Littman 1994; =-=Hu & Wellman 1998-=-; Brafman & Tennenholtz 2002; Bowling & Veloso 2001; Greenwald, Hall, & Serrano 2002)) as models of such interactions. The literature on learning in games in game theory (Fudenberg & Levine 1998) is m... |

236 | R-m.ax - A General Polynomial Time Algorithm for Near-Optimal Reinforcement Learning
- Brafman, Tennenholtz
- 2003
(Show Context)
Citation Context ...nberg & Levine 1998). Much of this work uses repeated games (e.g. (Claus & Boutilier 1997; Kalai & Lehrer 1993; Conitzer & Sandholm 2003)) and stochastic games (e.g. (Littman 1994; Hu & Wellman 1998; =-=Brafman & Tennenholtz 2002-=-; Bowling & Veloso 2001; Greenwald, Hall, & Serrano 2002)) as models of such interactions. The literature on learning in games in game theory (Fudenberg & Levine 1998) is mainly concerned with the und... |

215 | Rational Learning leads to Nash Equilibrium
- Kalain, Lehrer
- 1993
(Show Context)
Citation Context ...ntelligence, and related fields for quite some time (Kaelbling, Littman, & Moore 1996; Erev & Roth 1998; Fudenberg & Levine 1998). Much of this work uses repeated games (e.g. (Claus & Boutilier 1997; =-=Kalai & Lehrer 1993-=-; Conitzer & Sandholm 2003)) and stochastic games (e.g. (Littman 1994; Hu & Wellman 1998; Brafman & Tennenholtz 2002; Bowling & Veloso 2001; Greenwald, Hall, & Serrano 2002)) as models of such interac... |

81 | AWESOME: A general multiagent learning algorithm that converges in self-play and learns a best response against a stationary opponents - Conitzer, Sandholm - 2007 |

56 | Correlated Q-learning - Greenwald, Hall - 2003 |

45 | M.: Efficient learning equilibrium
- Brafman, Tennenholtz
- 2004
(Show Context)
Citation Context ...dominant notion in game theory and economics. It is only natural that similar requirements will be required from the learning algorithms. In order to address the above issue, Brafman and Tennenholtz (=-=Brafman & Tennenholtz 2004-=-) introduced the notion of Efficient Learning Equilibrium [ELE]. In this paper we deal with an improved version of ELE, where the agents’ surplus as a result of the learning process is required to be ... |

45 | Bundling equilibrium in combinatorial auctions
- Holzman, Kfir-Dahav, et al.
(Show Context)
Citation Context ...ning algorithms are in fact strategies in the corresponding (repeated) game, we in fact require that the learning algorithms will be an ex-post equilibrium in a (repeated) game in informational form (=-=Holzman et al. 2004-=-). The definition of OELE is of lesser interest if we cannot provide interesting and general settings where OELE exists. By adapting the results of (Brafman & Tennenholtz 2004) to the context of OELE ... |

4 |
Rational and covergent learning in stochastic games
- Bowling, Veloso
- 2001
(Show Context)
Citation Context ...f this work uses repeated games (e.g. (Claus & Boutilier 1997; Kalai & Lehrer 1993; Conitzer & Sandholm 2003)) and stochastic games (e.g. (Littman 1994; Hu & Wellman 1998; Brafman & Tennenholtz 2002; =-=Bowling & Veloso 2001-=-; Greenwald, Hall, & Serrano 2002)) as models of such interactions. The literature on learning in games in game theory (Fudenberg & Levine 1998) is mainly concerned with the understanding of learning ... |

2 | Reinforcement learning: A survey. JAIR 4:237--285 - Kaelbling, Littman, et al. - 1996 |