Probabilistic modeling and numerical simulation of neural circuits for multisensory integration

Abstract

People in real life receive stimulus information through various senses, and the process by which the brain integrates this information is called multisensory integration. Multisensory integration is an important branch of neuroscience, and the research on its neural mechanism holds significant application value to the development of artificial intelligence such as designing intelligent robots. Researches suggests that the brain likely employs Bayesian rules to integrate information and make judgments. In machine learning, neural networks based on Spike-Timing-Dependent Plasticity (STDP) have shown promising results in multimodal emotion recognition. In this paper, we model a neural network based on STDP, try to explain spike events using a probabilistic model, and unify network output with Bayesian calculation. This paper uses numerical simulation to verify the performance of the proposed network in multisensory classification problems. The results show that multisensory integration can improve classification accuracy and is better than the popular supervised learning method.