神经随机汇池网络的信息传递研究

发布时间：2018-05-13 20:33

本文选题：随机汇池网络 + 平均互信息　；参考：《青岛大学》2017年硕士论文

【摘要】：本文主要以平均互信息量和刺激特定信息为评价指标,以确定性非周期信号和语音信号为输入信号,在Gamma噪声和高斯噪声存在的环境下,研究饱和性突触型模型和积分发放型模型构成的随机汇池网络的信号传输功能。在饱和型突触神经的随机汇池网络中,把非周期信号作为输入的信号,用Gamma噪声模拟神经元细胞群体的内部噪声,把平均互信息以及刺激特定信息作为衡量指标,分别对兴奋性突触神经元构成的同质类随机汇池网络,以及兴奋性突触神经元与抑制性突触神经元共同构成的异质类随机汇池网络的随机共振现象进行了深入的研究,分析了网络内部的多种噪声源给平均互信息与刺激特定信息带来的变化;在积分发放神经随机汇池网络中,采用语音信号作为输入信号,内部噪声为Gamma噪声和高斯噪声这两种噪声,同样地,利用平均互信息和刺激特定信息评价,观察在积分发放神经随机汇池网络中的随机共振现象,分析改变噪声强度、积分发放神经元并联数目使随机汇池网络产生的信号传递效果;最后数值模拟结果表明,噪声可以增强随机汇池网络中的输入信号与输出信号的平均互信息,而刺激特定信息量可以更加详细地呈现出输入信号中各个分量编码的功效和神经元内部的噪声可利用性。本课题的研究结论具有十分积极的影响,尤其对于今后处理神经系统的信息具有一定的价值。
[Abstract]:In this paper, the average mutual information and specific stimulus information are taken as evaluation indexes, deterministic aperiodic signals and speech signals are used as input signals, and in the presence of Gamma noise and Gao Si noise, The signal transmission function of a random sink network composed of a saturated synaptic model and an integral distribution model is studied. In the random pool network of saturated synaptic nerve, the aperiodic signal is used as the input signal, the Gamma noise is used to simulate the internal noise of the neuronal cell population, and the average mutual information and the specific stimulation information are used as the measurement index. The Stochastic Resonance (SR) phenomena of the homogeneous random pool network formed by excitatory synaptic neurons and the heterogeneous random sink networks composed of excitatory synaptic neurons and inhibitory synaptic neurons were studied. The variation of average mutual information and stimulation specific information caused by various noise sources in the network is analyzed, and the speech signal is used as the input signal in the integral-distributed neural random sink network. The internal noise is Gamma noise and Gao Si noise. In the same way, using the mean mutual information and stimulating specific information evaluation, the stochastic resonance phenomenon in the integrated distributed neural stochastic sink network is observed, and the noise intensity is analyzed. The results of numerical simulation show that the noise can enhance the average mutual information between the input and output signals in the random sink network. Stimulation of specific amount of information can show in more detail the efficiency of each component of the input signal coding and the availability of noise inside the neuron. The conclusion of this study has a very positive effect, especially for the future processing of nervous system information.
【学位授予单位】：青岛大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R338

【参考文献】