Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster B in Poster Session B: Tuesday, August 4, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms

Representational Geometry of Optic Flow Determines Heading Direction Encoding in Macaque MSTd

Nathaniel Powell¹, Mary Hayhoe¹, Gregory C DeAngelis², Xue-Xin Wei¹; ¹University of Texas at Austin, ²University of Rochester

Presenter: Nathaniel Powell

Efficient-coding theories predict that more probable stimuli should receive greater neural resources and higher Fisher Information. However, macaque MSTd appears to violate this prediction: many neurons prefer lateral headings, whereas population sensitivity is highest near forward and backward headings. We show that this dissociation follows from the asymmetric geometry of the optic-flow manifold. In a hierarchical model in which heading generates optic flow, MT-like units encode local motion, and MST-like units pool MT responses, lateral headings are over-represented because they occupy compressed regions of the manifold, whereas forward/backward headings support higher Fisher Information because they lie along stretched regions. Reanalysis of visually responsive MSTd neurons showed the same qualitative geometry, tuning-density asymmetry, and Fisher Information profile. A simplified elliptical-manifold model reproduced the same relationship between representational allocation and sensitivity. These results suggest that MSTd heading tuning reflects the geometry of the sensory input manifold, not stimulus probability alone.

Topic Area: Computational Models of Vision & Visual Cortex