Timezone: »

 
Regret Decomposition in Sequential Games with Convex Action Spaces and Losses
Gabriele Farina

Fri Dec 07 12:30 PM -- 12:50 PM (PST) @
in Smooth Games Optimization and Machine Learning »
Event URL: https://sgo-workshop.github.io/papers/12/sgo.pdf »

We derive a new framework for regret minimization on sequential decision problems and extensive-form games with general compact convex sets at each decision point and general convex losses, as opposed to prior work which has been for simplex decision points and linear losses. We call our framework laminar regret decomposition. It generalizes the CFR algorithm to this more general setting. Furthermore, our framework enables a new proof of CFR even in the known setting, which is derived from a perspective of decomposing polytope regret, thereby leading to an arguably simpler interpretation of the algorithm. Our generalization to convex compact sets and convex losses allows us to develop new algorithms for several problems: regularized sequential decision making, regularized Nash equilibria in extensive-form games, and computing approximate extensive-form perfect equilibria. Our generalization also leads to the first regret-minimization algorithm for computing reduced-normal-form quantal response equilibria based on minimizing local regrets.

Author Information

Gabriele Farina (Carnegie Mellon University)

More from the Same Authors