A Three-Neuron Model of Information Processing During Bayesian Foraging

Abstract

A foraging animal is often confronted with uncertainty of resource abundance. A Bayesian model provides the optimal forgaing policy when food occurrence is patchy. The solution of the Bayesian foraging policy requires elaborate calculations and it is unclear to what extent the policy could be implemented in a neural system. Here we suggest a network architecture of three neurones that approximately can perform an optimal Bayesian foraging policy. It remains to be shown how the network could be self-learned e.g. through Hebbian learning, and how close to to the optimal policy it can perform.