NeurIPS 2020

No-Regret Learning Dynamics for Extensive-Form Correlated Equilibrium

Meta Review

All the reviewers agreed that the paper makes a significant and novel contribution. A decentralized regret minimization algorithm that converges to EFCEs is an important contribution to the field. There were some minor issues with presentation that were cleared by the author response and discussion. The connection to Pure CFR by Gibson should be referenced, and the paper should be explicitly stating the limitation that computing the EFCE requires storing every generated strategy. Please address these in the final version.