Uncertainty and learning in a biological ring attractor

Abstract

Navigation often occurs under uncertain conditions, where spatial information is incomplete. In this thesis project, we set out to study how uncertainty affects the head direction (HD) system in the Drosophila brain, a biological ring attractor where a bump of activity tracks the animals HD as it moves around. We found that the shape of this bump of activity reflects the uncertainty of the environment, and correlates with adaptive remapping of the heading representation. Specifically, the bump becomes wider in uncertain conditions where HD encoding is unstable. Notably, individuals with wider bumps are better able to learn a dramatic remapping of the HD representation when the environment changes. We found that an environment with multiple sensory cues improves HD encoding and decreases bump width. In a multi-cue environment, different individuals assign different levels of uncertainty to different cues, and this predicts the relative weight that they will assign to each cue during cue conflict. Together, these results show how uncertainty can alter the format of a spatial map to produce adaptive changes in cue integration and learning.