Skip to main content
Menu

Aldo Pacchiano Seminar

AIMS Seminar - Friday 25th November

On the Statistical Complexity of Batch Learning with Function Approximation: Theory and

Bio: Aldo Pacchiano received a bachelors and master’s degrees from MIT and Cambridge University, and his PhD from UC Berkeley where he was co-advised by Prof. Peter Bartlett and Prof. Michael Jordan. He is currently a postdoctoral researcher at Microsoft Research NYC. His research work and interests span both theoretical topics in online learning, bandits, and reinforcement learning (RL), and practical questions in deep RL, fairness, and experiment design. His theoretical work has dwelled on fleshing out off-the-current-trodden-path learning models for bandits and RL, transfer learning in RL problems, as well as advancing the state of the art on the problem of model selection for bandits and RL. His applied work has had a particular focus on designing theoretically sound RL algorithms that can be implemented in large scale practical settings as well as devising algorithms to bring the power of online learning to improve the ways in which machine learning is used in scientific experimentation. Outside of research, he enjoys creative writing.