Humans and animals coordinate behavior across many spatial and temporal scales, from rapid movements to long-term goals and habits. How the brain links these different levels of behavior remains poorly understood.
The Action Computation Lab investigates how distributed brain circuits learn representations that connect perception, action, and internal goals, to support multiscale control and policy learning. We develop theoretical and computational models and test them in large-scale neural and behavioral datasets.
Our work aims to uncover general principles of hierarchical representation, multi-area computation, and embodied learning, while also inspiring new approaches to flexible and continual learning in artificial systems.