"Foundations for Multi-Agent Learning" – Maxwell Fishelson, Talks at TTIC

“Theoretical Foundations for Multi-Agent Learning” Maxwell Fishelson, MIT Originally recorded on March 30, 2026, at TTIC. In this talk, Maxwell Fishelson explores the theoretical foundations needed for learning in multi-agent, dynamic environments where data is no longer independent and identically distributed. He presents advances in regret minimization and calibration, showing how algorithms can achieve robust performance and trustworthy forecasting even in the presence of strategic interactions. Timestamps: 00:00 Introduction 01:35 Talk begins 50:20 Q&A #MultiAgentSystems #GameTheory #MachineLearning #AI #RegretMinimization #Calibration #LearningTheory #Algorithms #Research #TTIC