Tom Schaul

Dalle Molle Institute for Artificial Intelligence Research

Papers

Total Citations

274

H-Index

About

Tom Schaul is a leading researcher in artificial intelligence, whose work has fundamentally advanced reinforcement learning (RL) through novel approaches to exploration and transfer learning. His key contributions lie in making RL agents more sample-efficient and adaptable. Schaul is best known for introducing **Successor Features for Transfer in Reinforcement Learning** (2016, 184 citations), a seminal framework that enables agents to generalize learned knowledge across tasks with different reward functions but identical dynamics. This work provides a principled method for zero-shot transfer, dramatically reducing the need for retraining. Earlier, he pioneered **parameter-based exploration** (2010, 79 citations), a technique that perturbs policy parameters directly rather than actions, unifying RL with black-box optimization and proving especially effective in high-dimensional control problems. Schaul also explored **artificial curiosity** for autonomous space exploration (2011, 11 citations), applying intrinsic motivation to drive agents toward novel states. His research has been instrumental in shaping modern deep RL, influencing algorithms used in robotics, game-playing, and autonomous systems. Schaul’s work continues to inspire students and researchers seeking to build more intelligent, adaptive agents.