Explore coordinate methods for solving matrix games in this 25-minute IEEE conference talk presented by researchers from Stanford University. Delve into bilinear saddle point problems, algorithms for matrix games, and the rationale behind using coordinate methods. Examine key technical challenges and learn from a specific matrix games example. Gain insights into the design of estimators, implementation strategies, and data structure interfaces. Investigate a simpler problem with zero drift before reaching the concluding remarks on this advanced topic in game theory and optimization.
Overview
Syllabus
Intro
Bilinear saddle point problems
Algorithms for matrix games
Why coordinate methods?
Key technical challenges
Example: matrix games
Design of estimators
Implementation
Data structure interface
A simpler problem: zero drift
Conclusion
Taught by
IEEE FOCS: Foundations of Computer Science