Neural Network Models in Evolutionary Ecology: How to Handle Optimal Reaction to Multiple Inputs

Masashi Kamo (Kyushu University)

01/01/22,16:00-17:30
at Room No.3521 (5th floor of the 3rd building of the Faculty of Sciences)



Central foraging has to solve simultaneously two contradicting problems; efficient foraging from an activity center (colony or roost) and efficient roosting to the center. Most theoretical models make light of the latter, and often simplify it as a straight- line motion of coming home. Organisms in nature are far from omniscient. They wander while coming home as well as foraging. I propose two types of models to predict colony/roost distribution of central foraging animals in variable resource environment. The first type is a spatial moving average model assuming omniscient foragers. This plays a role of a null model for rejecting omniscience. The second one is a configuration i-sate individual-based model (cIBM) using Random Urn Model (RUM) as a learning system. I apply the two models to data of bumble bees and egrets and estimate their colony distribution.