Modeling sensory-motor decisions in natural behavior.

摘要: Although a standard reinforcement learning model can capture many aspects of reward-seeking behaviors, it may not be practical for modeling human natural behaviors because the richness dynamic environments and limitations in cognitive resources. We propose modular that addresses these factors. Based on this model, inverse algorithm is developed to estimate both rewards discount factors from behavioral data, which allows predictions navigation virtual reality with high accuracy across different subjects tasks. Complex trajectories novel reproduced by an artificial agent based model. This provides strategy estimating subjective value actions how they influence sensory-motor decisions behavior.