Information Transmission Characteristics Of Polychronous Group In Izhikevich Neural Network Research

Izhikevich neural network is a relatively new neural network, which can generate a large number of polychronous groups during training by using Izhikevich simple model of spiking neuron and spike-timing-dependent plasticity(STDP). Polychronous group can be used as a concrete example of group of neurons, and can analyze the information of the receptor neurons receiving from the anchor ones quantitatively, as well as the characteristics of information transmission from the source to the target area. In a word, polychronous group is of great significance in further studying the biological signal encoding and transmission. In order to research the characteristics of information transmission in polychronous group of Izhikevich neural network, we solved four problems, that is, how to discriminate the dynamic stability of polychronous group, how to design stimulus to make neurons excitatory in group, how to obtain the information transmission tendency and characteristics of neurons in group, and how to get neurons’ancestor ones.In order to discriminate whether the polychronous group reaches dynamic stability, a discrimination method of bimodal distribution is proposed which is based on neurons whose synaptic weights are initialized to Gaussion distribution and random distribution. In addition to initial synaptic weights, the most important step is adjusting the learning time constantly to decide when to reach the ideal state. For the neural network, this method solves the disadvantages of training a fixed time and saves time for further study.To make the excitatory times of neurons in the polychronous groups more dense, a stimuli-forming algorithm is raised. Not only the groups are more active, but the correlation measuring of neurons is more correct basd on the intensive times.With the purpose of obtaining information transmission trend of neurons in a polychronous group, a method of measuring correlation between neurons by using van Rossum’s Distance is presented.In addition, neurons’amount of information in the polychronous group is not only affected by the ancestor neurons within the group, also affected by the ancestor neurons in the overall Izhikevich neural network. So we present an algorithm named ancestor-neurons-getting as to verify the rationality of information transmission characteristic of neurons in the group. Besides, the number of effective ancestor neurons in the network can be used to analyze the amount of information.Characteristics of information transmission can be divided into two categories via studying Izhikevich neural network, that is the ending-neuron is an inhibitory one or an excitatory one. For the former, mother neuron and middle neurons can receive abundant information from anchor ones, and the ending-neuron receives little information. For the latter, the ending-neuron can receive rich information as well as the mother neuron and middle neurons.