Talk: Tuning movement to optimize information harvesting
Speaker: Malcolm MacIver, Northwestern University
Title: Tuning movement to optimize information harvesting
Emcee: Nidhi Seethapathi
Backend host: Elnaz Alikarami
Details: https://neuromatch.io/abstract?submis...
Twitter: @malcolmmaciver
Presented during Neuromatch Conference 3.0, Oct 26-30, 2020.
Summary: Information about potential fitness enhancers is dispersed in space, which provides a logic to the heterotroph strategy of moving to acquire resources. But how ought animals move to optimally acquire information? Current work on Lévy-flight like trajectories may apply to the context of acquiring resources that are far outside the range of sensory systems, but are not predictive when resources are in range. Entropy minimization approaches like infotaxis provide predictions for how animals ought to move when targets are in sensory range, but these do not agree well with measured trajectories. We present a new approach, energy-constrained proportional betting, and show that it generates trajectories that agree well with measured trajectories of animals localizing targets across four species spanning insects, fish, and mammals. In addition to predicting sense organ movements, the method also performs well for prescribing sensor locations in mobile robots in weak and noisy signal conditions.