Abstracts

Most events on this list can be tied to a specific page somewhere on the internet.  List entries contain links to such pages through the "Details".   Occasionally events have no such page or the information on it is inadequate.  In such cases I attempt to cobble together a details page myself.



From: Brian Xie >
Subject: Re: Poker AI Talk
Date: August 30, 2017 at 9:11:45 AM EDT

Title: Strong Game-Theoretic Strategies: Beyond Two Agents
Speaker: Sam Ganzfried, Florida International University (School of Computing and Information Sciences)
Time: 4:00 PM to 5:00 PM
Refreshments: Pizza
Location: 4-370
Contact: Brian Xie, brianxie@mit.edu
Relevant URL: http://www.ganzfriedresearch.com/

On September 15, Prof. Sam Ganzfried (CMU) will be giving a talk "Strong game- theoretic strategies beyond two agents" about his research on Poker AI agents. His poker agents have been tremendously successful and have achieved state-of- the-art performance (including Claudico, which participated in the first Brains vs AI competition in 2015). More about his research on his website: http://www.ganzfriedresearch.com/

He is still actively working in the area, and is looking to collaborate with MIT students and faculty. He has mostly worked on two-player poker variants, but is extending his research to multi-player agents. He is planning to submit for both the 2-player and 6-player variants for the 2018 Computer Poker Competition.

This talk will give MIT students/faculty a great opportunity to collaborate with one of the most prominent people in the intersection of AI and game theory.

Brian Xie
brianxie@mit.edu
Abstract:
Efficient algorithms have been developed for approximating Nash equilibrium strategies in two-player zero-sum games, even extremely large ones. In fact, just this year agents for two-player no-limit Texas hold 'em which has around 10^165 nodes in its game tree were created by the University of Alberta and Carnegie Mellon University that defeated human professional players. Creating agents for games with more than two agents is significantly harder, both from computational as well as conceptual perspectives. For two-player zero-sum games like two-player poker a Nash equilibrium strategy would be "unbeatable" in expectation, while one would have no guarantee at all for more agents. Furthermore, from a complexity perspective an equilibrium can be computed in polynomial time in two-player zero-sum games, while it is PPAD-hard for more agents. Nonetheless, many of the techniques developed in the course of research on two-player agents can be applied for more agents, though theoretical guarantees are more limited.

Bio:
Sam Ganzfried is an Assistant Professor in the School of Computing and Information Sciences at Florida International University in Miami. He received a PhD in computer science from Carnegie Mellon University in 2015 for his dissertation "Computing Strong Game-Theoretic Strategies and Exploiting Suboptimal Opponents in Large Games" and holds an A.B. in math from Harvard University. His research interests include artificial intelligence, game theory, multiagent systems, multiagent learning, large-scale optimization, large-scale data analysis and analytics, and knowledge representation. He created two-player no-limit Texas hold ‘em agent Tartanian7 that won the 2014 Annual Computer Poker Competition and Claudico that competed in the inaugural Brains vs. Artificial Intelligence competition in 2015 against the strongest human specialists in the world for that poker variant. He organized the AAAI Workshop on Computer Poker and Imperfect Information in 2014 and 2015, as well as the First Tutorial on Computer Poker at the 2016 Conference on Economics and Computation and again at AAAI in 2017. He has developed new courses in Game Theory and Artificial Intelligence at FIU.