 |
Center for Nonlinear Studies |
In many real-world strategic scenarios, humans do not alternate actions but instead act and also react to the actions of other players as well as changes in the social and physical environment. Traditional game theory typically fixes the action sequence of players and the environment *a priori* and does not allow for the action-reaction dynamics of real-world scenarios. While developments in the field of stochastic games attempt to assuage this shortcoming, its structure precludes the analysis of complex environments and consequently is only useful or presenting high level abstractions. In the presentation, I will outline a new modeling framework created by Wolpert et al. known as "event driven game theory" that combines stochastic process theory with game theory to facilitate a more accurate representation of real world strategic scenarios. To illustrate its applicability, I will show how event given game theory can be used to improve attack detection in cyber-networks and will present some preliminary results.
Host: Joshua Neil