Machine Decision and Game Theory
Description
Automated systems that make decisions (e.g, robots, programmed traders, medical decision aids) face many of the same tradeoffs as humans when making decisions in the context of limited resources, risk, and uncertainty. Bayesian decision theory and game-theoretic models that originated with John von Neumann and Oskar Morgenstern can be used to analyze such systems.
At Brown, we are working towards extending the classic theories to handle more complex forms of automated reasoning: e.g., machine learning to compensate for and adapt to the behavior of adversaries and collaborators alike. We are also exploring theories of machine decision-making similar to Herb Simon's and Daniel Kahneman and Amos Tversky's theories about how humans depart from the classic utility-maximizing model.
We apply machine decision and game theory to develop algorithms that animate robots, control disembodied robots (i.e., software agents) that negotiate and trade on the web, and predict the performance of complex financial instruments in volatile markets, just to name a few of the myriad of applications. The research projects associated with this theme typically include a mix of theoretical and empirical approaches.
Faculty
| Amy Greenwald |
Topics or Projects
| Planning Under Uncertainty |
| Multiagent Learning in Games |
| Reinforcement Learning in Markov Games |
| No-Regret Learning and Games |
| Page Owner: Webmaster | Last Modified: Mon Oct 23 11:47:20 2006 |