A reinforcement learning model for macaque monkey tool-use

摘要

Iriki et.al. have shown that although Japanese monkeys took about two weeks to be able to use a rake to get distant food, they could immediately use two rakes to get more distant food. They also found neurons in caudal postcentral gyrus, whose visual receptive field is enlarged with using a rake. This study presents a model based on the reinforcement learning theory in order to explain the monkeys' behavior. [MODEL] This model is based on a hypothesis that monkeys can evaluate tools according to their distance to food. The model consists of a conventional Actor-Critic model and a tool-evaluation-component. [RESULTS] Computer simulation reproduced the above experimental data. This verifies the hypothesis that monkeys can evaluate tools.