Identification Method of Time-varying Neural Dynamics System Based on Chebyshev Polynomial Expansion

Abstract

The invention proposes a time-varying neurodynamic system identification method based on Chebyshev polynomial expansion. This method first uses Volterra series to characterize the time-varying nervous system composed of simulated input and output spike sequences, and expresses the feedforward and feedback kernel functions with different Volterra kernels; then, uses Laguerre basis functions to expand the time-varying Volterra kernel , to obtain the time-varying generalized Laguerre–Volterra model; then, use Chebyshev polynomials to expand the time-varying parameters of the time-varying generalized Laguerre–Volterra model, and transform the time-varying model into a time-invariant model; finally, use the forward orthogonal The regression algorithm selects meaningful model items, and uses the generalized linear fitting algorithm to estimate the time-invariant parameters, and then obtains the time-varying parameters and the original time-varying kernel function through reverse solution. Compared with the existing adaptive filtering technology, this method has better tracking ability for strong non-stationary nervous system signals, can realize accurate tracking of time-varying system kernel functions, and model neural systems, especially for massive high The systematic modeling of three-dimensional data provides a new research method, which is of great significance for revealing the complex neural dynamics mechanism of the brain's information processing.

Description

technical field [0001] The invention provides a time-varying neurodynamic system identification algorithm based on Chebyshev polynomial expansion, which provides a new analysis method for time-varying system identification facing spike sequence signals, and belongs to the field of system identification. Background technique [0002] The nervous system is a dynamic system, and the underlying mechanism of neuron spike activity exhibits time-varying characteristics. This time-varying may be extremely slow, but its changes cannot be ignored as time accumulates. Therefore, using a time-invariant model to analyze the potential mechanism of neuron spiking activity obviously cannot obtain long-term stable and reliable results. Analyzing the potential time-varying laws of neurons, and developing time-varying system modeling and identification applications of neuron spike sequences have gradually attracted the attention of researchers. [0003] Most of the time-varying system modelin...

AI Technical Summary

Problems solved by technology

However, if the parameters of the time-varying system change too fast, the estimation of the parameters of the time-varying system will be delayed due to the convergence defect of the adaptive algorithm.

Images