Axes de recherche du CREUSET- Game theory : networks, complexity and dynamics.
report (pdf)
Supervisor : Philippe SOLAL
Members: Richard BARON, , Pascal BILLAND, Christophe BRAVARD, Jacques DURIEU, Philippe SOLAL
PhD students : Amandine GHINTRAN
Associated : Sylvain BEAL
The main interest of each member of this team is game theory with applications to the formation of networks. In network formation games, players are represented by nodes in the network, and their relationships by links between nodes.The emphasis is on the understanding of how such networks form and evolve over time. Another interest of some members of the team is the computational aspects of game theory and its implications. Computational complexity studies algorithmic issues involved in deciding in a reasonable amount of time whether a game admits a given property.
In addition to its importance in computer science, computational complexity provides one of the mainmotivations for theories of bounded rationality in economics. Perhaps the most general definition of bounded rationality is rationality exhibited by decision makers with cognitive limited abilities, limitations of both knowledge and computational capacity. Theories of bounded rationality aim to model how players would proceed in situations in which deciding how to coordinate independent choices is too time consuming. Recent trends in game theory study adaptive learning models in which players follow simple decision rules. Decisions taken on the basis of myopic best responses, reinforcement rules, imitation, or other short-sighted rules, coalesce in the long-run into limit sets or conventions. Another way to model bounded rationality within the context of repeated games is to suppose that players are restrictedto using strategies which are implementable by a machine (a finite automaton, a perceptron or a Turing machine for example). The way in which players are restricted and how the set of machines are ordered by the players are two aspects which have a non negligeable effect on the set of Nash equilibria of the repeated game. Evolutionary game theory and machine games are two attractive topics that members of the team investigate.
The main interest of each member of this group is game theory and its applications to the formation of networks. In network formation games, players are represented by nodes in the network, and their relationships by links between nodes. The emphasis is on the understanding of how such networks form and evolve over time.Another interest of some members of the group is the computational aspects of game theory and its implications. Computational complexity studies the algorithmic issues involved in deciding in a reasonable amount of time whether a game admits a given property.In addition to its importance in computer science, computational complexity provides one of the main motivations for theories of bounded rationality in economics. Perhaps the most general definition of bounded rationality is that which is exhibited by decision makers with cognitive limited abilities, limitations of both knowledge and computational capacity. Theories of bounded rationality aim to model how players would proceed in situations in which deciding how to coordinate independent choices is too time consuming.Recent trends in game theory study adaptative learning models in which players follow simple decision rules. Decisions taken on the basis of myopic best responses, reinforcement rules, imitation, or other short-sighted rules, coalesce in the long-run into limit sets or conventions. Another way to model bounded rationality within the context of repeated games is to suppose that players are restricted to using strategies which are implementable by a machine (a finite automaton, a perceptron or a Turing machine). The way in which players are restricted and how the set of machines are ordered by the players are two aspects which have a non-negligible effect on the set of Nash equilibria of the repeated game.Evolutionary game theory and machine games are two attractive topics that members of the group investigate.
Théorie des jeux : dynamique et réseaux
L'intérêt principal de chaque membre de cette équipe est la théorie des jeux avec des applications à la formation des réseaux. Dans des jeux de formation de réseaux, les joueurs sont représentés par les noeuds du réseau et leurs connexions par les arêtes entre les noeuds. La problématique est centrée sur la compréhension de la façon dont de tels réseaux se forment et évoluent avec le temps.
Un autre intérêt de certains membres de l'équipe porte sur les aspects computationnels de la théorie des jeux. La complexité algorithmique s'intéresse à la question du temps de calcul nécessaire pour décider si un jeu admet une propriété donnée. En plus de son importance en informatique, la complexité algorithmique fournit une des motivations pour développer des modèles de rationalité limitée dans les sciences économiques. Sans doute la définition la plus générale de la rationalité limitée est celle de décideurs ayant des capacités cognitives limitées, avec des limites portant sur les connaissances et sur les capacités calculatoires.
L'objectif des théories de la rationalité limitée est de modéliser comment les joueurs procéderaient dans des situations où la coordination des choix individuels prend trop de temps. Des avancées récentes en théorie des jeux étudient des modèles d'apprentissage adaptatif dans lesquels les joueurs suivent des règles de décision simples. Les décisions prises sur la base de règles de réponses optimales myopes, de renforcement, d'imitation, ou d'autres règles myopes, convergent dans le long terme vers des conventions.
Une autre manière de modéliser la rationalité limitée dans le contexte de jeux répétés est de supposer que les joueurs sont contraints par l'utilisation de stratégies qui sont implémentables par une machine (un automate fini, un perceptron ou une machine de Turing par exemple). Les jeux évolutionnaires et les jeux de machine sont deux thèmes que les membres de l'équipe étudient
