Information-Theoretic Gradient Flows in Mouse Visual Cortex

Erik D. Fagerholm^1*Hirokazu Tanaka²Milan Brázdil¹

• ¹Masarykova univerzita, Brno, Czechia
• ²Tokyo Toshi Daigaku, Setagaya, Japan

Neural activity patterns can be viewed as probability distributions that change over time. Modelling the ways in which these distributions change as they are progressively transmitted among neural regions is central to understanding how information is organized in the cortex. We introduce a framework that describes these transformations as information-theoretic gradient 'flows' — dynamical trajectories that follow the steepest ascent of entropy and expectation. The relative strengths of these two functionals provide interpretable measures of how distributions are continually reshaped as they propagate along the networks of the brain. Following construct validation in silico, we apply our methodology to publicly available continuous ΔF/F two-photon recordings from the mouse visual cortex. We identify prominent bi-directional transformations between the rostrolateral area and the primary visual cortex across all five mice. Our approach offers a generalizable method that can be used to separate information-theoretic flows, not just among brain regions, but also across neuroimaging modalities and observational scales.