Nauka i Obrazovanie (Jan 2014)
Neuron Network Simulation Taking into Consideration the Biological Properties of Neuron
Abstract
The simulation of the neural network element based on the properties of the neuron as a biological object and principles of the recording and processing of information in neurons was done. In contrast to the simplified representation of a neuron as the logic element in a network data processing a neuron is considered as an analog element to store and generate information. In neural information processing, dendrites, synapses, and axons are involved to form and transmit pulses. Pulses generated in the neuron are considered. Expressions for the shape and spectrum of pulses and their FM sequence were written. The pulses are generated from the thermal white noise, and it is possible to find the transfer function of individual dendrites. Due to these functions the transfer function of the neuron as a whole is recorded. Neuron driven by external signals performs addition to the different time delays of pulses coming from different dendrites and with different repetition periods. The functional diagram of the neuron is described. The equation for the vector of signals of several separate layer neurons is obtained. Signals generated in each layer are not a set of ones and zeros, but pulse frequency modulated signals which control synapses of next layer. The principle of control synapses using them arriving on FM pulse sequences is described. The results of modeling the mechanism of allocation synapse FM modulation of the supplied thereto pulse signal were carried. The higher is pulse repetition frequency in the signal, the more molecules of neurotransmitter released into the synapse, and the higher is the potential in a controlled dendrite, which determines the frequency of pulse generation. The paper considers a model of a neuron that can be a basis for the creation of artificial intelligent information processing systems.
Keywords
