3:00 pm Wednesday, November 8, 2017
Random Structures Seminar: Evaluating structured Perfect Bayesian equilibria in dynamic games with asymmetric information by Deepanshu Vasal (UT Austin) in RLM 8.136
Many real-world scenarios involve dynamic, strategic interaction of many decision makers with different sets of observations. This could be people in a social network, firms in a market, or even species in an evolutionary sense. In this talk, I will consider a general abstract model of a class of dynamic games with asymmetric information, where there are N players with conditionally independent (Markovian) private types, who publicly observe actions of everybody else. An appropriate notion of equilibria for such games is Perfect Bayesian equilibrium which involves a strategy and belief profile which are coupled together across time, which makes finding such equilibria hard to compute. For the general class of games considered, I will present a novel methodology to sequentially decompose this problem across time, in an analogous way dynamic programming sequentially decomposes a dynamic optimization problem, such that the complexity of the problem reduces from double exponential to linear in time. I will present a running public goods example to illustrate the methodology to compute signaling equilibria of that game. Submitted by
|
|