Research on algorithms of neural network based on self-growing of neuron

Abstract

In this paper, a growth algorithm of neural network based on self-growing of neuron is proposed. Firstly, the self-growth model of neuron is established based on Newton's gravitational force and Brownian motion. Then, the constraint function of neuron axon growth is established by gravitational coefficient and Brownian coefficient. Then, the algorithm of neural network is established by using the breadth-first search algorithm and the discrete-variable optimization algorithm. Finally, the numerical simulation results show that the algorithm can effectively simulate the growth and development of neural networks from single neuron growth to multiple neurons. It also verifies the feasibility of using the algorithm to realize the rhythm motion control of quadruped robot.