突触神经随机汇池网络的信息传递研究

Study of Information Transmission in Synaptic Neural Stochastic Pooling Network

针对突触神经网络的信息传输问题,本文研究了饱和性突触神经随机汇池网络的信息传递特性,分析了网络内部噪声源对于互信息和刺激特定信息的影响,给出了互信息和刺激特定信息的数值计算方法,并对刺激特定信息分布进行实验研究。实验结果表明,在兴奋性和抑制性突触神经共同组成的异质类随机汇池网络中,存在适量噪声能够增强网络信息传递的随机共振现象,远离饱和区的信号幅值对神经元群体信息传递效率的贡献较大。本文的研究结果对进一步理解突触神经信号处理机制具有重要指导意义。

This paper focuses on the information transmission in synaptic neural networks.We study the information transfer characteristics of saturating synaptic neuron stochastic pooling networks,and analyze the effects of network internal multi-level noise sources on the mutual information and the stimulus-specific information.We present the numerical algorithms for calculating the mutual information and the stimulus-specific information.The results demonstrate the noise-enhanced effect of information transmission in heterogeneous stochastic pooling networks of multi-synaptic excitatory and inhibitory pathways,and the signal amplitude that can suffer from saturation contribute significantly to the effect of information transmission of groups of neurons.These research results are of great significance for the synaptic neural signal processing mechanism that deserves to be further understood.