Game Theory studies the phenomena occurring when independent, autonomous entities, called agents or users, act selfishly; game theoretic techniques are now being used to model and analyze networks.
Dr. George Karakostas , (McMaster University)
Game Theory studies the phenomena occurring when independent, autonomous entities, called agents or users, act selfishly; game theoretic techniques are now being used to model and analyze networks. This project aims to develop a more realistic modelling of communication and data networks of selfish users using game-theoretic models, study the effects that selfish behaviour has on the overall network performance, and the designs of networks which prevent the rapid degradation of the performance due to such behaviour.
Achievements in the past year include the development of a model for a natural notion of malicious behaviour in a selfish routing network which extended the results of the team’s previous work on the resulting price of anarchy for linear latency function to general latency functions. In addition, the team showed the existence of optimal taxes (taxes that force selfish users in a selfish routing network) even in the case of elastic demands, i.e., the flow rates for every Origin-Destination pair depend on the path costs experienced by the users. The result holds for very general demand functions and its proof relies heavily on fixed-point and mathematical programming duality theorems.