Dharshan Kumaran

Training a network only to predict its visual input makes head direction, boundary vector, place and egocentric boundary cell responses appear, and lets agents reuse behaviours in unfamiliar environments.

The paper uses fMRI to study human navigation behavior in dynamic environments, finding specific brain activity patterns associated with back-tracking, detours, and shortcuts.

The paper proposes a new theory of reward-based learning where the prefrontal cortex functions as a meta-reinforcement learning system trained by the dopamine system.