Ryan D'Orazio

TL;DR: This work develops and advocate for this hindsight rationality framing of learning in general sequential decision-making settings, and re-examines mediated equilibrium and deviation types in extensive-form games, thereby gaining a more complete understanding and resolving past misconceptions.

TL;DR: This work derives approximation error-aware regret bounds for $(\Phi, f)$-regret matching, which applies to a general class of link functions and regret objectives and provides a theoretical justification for RCFR implementations with alternative policy parameterizations, including softmax.

TL;DR: In this paper, the authors formalize behavioral deviations as a general class of deviations that respect the structure of extensive-form games, and introduce an extensive form regret minimization (EFR) algorithm that achieves hindsight rationality for any given set of behavioral deviations with computation that scales closely with the complexity of the set.

TL;DR: In this article, the authors derive approximation error-aware regret bounds for (¶hi, ƒ)-regret matching, which applies to a general class of link functions and regret objectives.