Exploration Of Bionic Artificial Neural Network Based On Self-growth

With the rise of intelligence,people's demand for strong artificial intelligence has become increasingly urgent.The disadvantages of deep neural network is inevitable in the realization of strong artificial intelligence.Therefore,more and more scholars are engaged in the study of bionic neural network,but the problem of current bionic neural network is that neurologists know too little about the working principle of biological neural network,let alone have the knowledge,learning,Innovative way to understand the principles of the neural network.Computing scientists want to have a strong artificial intelligence through a complete biological brain simulation,there are still very difficult.So,a new bionic neural network implementation method is needed,and it should not be studied only from the microcosmic behavior of the biological neural network,but from the perspective of chemistry to self-organization and self-growing neural networks.It is clear that we are not trying to reproduce a biological neural network,but to study how to make independent neurons can be like a biological neuron network in the case of external stimuli to achieve self-organization,self-growth and self-learning.According to the survey,the biological neural network is a sca le-free network,and the scale-free network generation method can be done in the mathematical method.In this paper,the growth process of neurons is modeled by the microscopic behavior of biological neuron cells and the generation method of scale-free network.On this basis,a self-growing neural network generation method is proposed.Because of the growth model established by the pure mathematical method,we cannot control the growth of the network structure without considering the position information of the nodes in the network.So we made a certain revision on the basis of the model,so that it can control the growth of neurons.In order to solve the neural network growth efficiency,we propose a number of solutions in the simulation process.In order to improve the network efficiency and guide the network generation function,we designed the bionic Drosophila compound eyes neural network based on the current simulation platform,and then we can explore the self-generation algorithm on the basis of the network.The network generation method proposed in this paper is verified on the simulation platform developed by the laboratory independently.The results show that the network generated by the network generation,learning and filtering algorithms proposed in this paper can distinguish and judge the different input of the network.In addition,the experiment also found that the network initialization parameters and network-generated infrastructure on the network generation efficiency and generation function has an important impact.On the whole,this article provides a way to explore new intelligent network structures from a bionic point of view.Since the work of this paper is still in the early stage of research in this field,there are still many need to further optimize and improve the space in the whole algorithm.