Research On Identification Of Parameters Of Neuron Model Based On Firing Characteristics

Research on information encoding mechanism of neural system is the key link in uncovering the functions of brain, and neural model analysis is an important method in learning neural information encoding.. However, the parameters and structures of current models can not describe all the neural firing activity discovered in electrophysiological experiments, and correcting model parameters based on the experiment or model data to make the model present expected firing characteristics becomes one clear priority in research on neuron modeling and analysis. Therefore, this paper proposes the identification method of neural model parameters based on neural firing features.Firstly, we put forward the identification of Back Propagation(BP) neural network method based on the neural firing trajectory.(1) we adopt BP neural network to identify parameters from the input and output data of single neural models, to make the identified BP neural network can present the transfer characteristics of models and precisely predict the firing trajectory.(2) we apply this method in the input and output data of experimental recordings, and make the network study the neural transfer characteristics and predict the firing trajectory at a certain accuracy.Secondly, we propose the estimation of parameters of Thalamic Cell(TC) model which are related with the Parkinson's disease(PD) based on PSO.(1) we put forward that, the relay responsibility of TC model can determine the PD state, and the relay responsibility is mainly related with two parameters: the inhibitory synaptic current from BG to TC and the maximum ionic conductance densities of calcium ion. So we conduct identification research on the two related parameters based on firing features including the relay responsibility of TC model.(2) Based on the PSO algorithm, we estimate the key parameters of TC model from the PD related firing features, and make the identified model present the same firing characteristics and effectively predict firing features.(3) With an arbitrarily given relay reliability value, we adopt PSO in estimation the model parameters, and obtain the expected firing features from the model.Finally, we adopt PSO algorithm in the identification of neural model parameters based on adaptive firing features of neurons.(1) We extract the spiking features involved with the adaptation from the model data to estimate model parameters, to make the identified model precisely predict the adaptive neural spikes. Then the effectiveness of this identification method in idol situation is proved.(2) We realize the identification on adaptive spiking features extracted from experimental data. For a adaptive model whose firing characteristics are close to experimental data, the model parameters are estimated, and the identified model can effectively predict the adaptive neural spikes corresponding to fluctuated current input.This paper conducts the research on neuron modeling and analysis from the perspective of system identification based on single neural firing features. It provides support for the further analysis and control of neural signal, and provides solid foundation for the exploration of neural signal encoding mechanism.