Formation Mechanism Of Neural Circuits Based On Temperature-dependent Neurons

With the rapid development of artificial intelligence,research on neural networks is getting deeper and more comprehensive.As the third-generation neural network,pulse neural networks are widely concerned.How to construct an intelligent network which is similar to the human brain neural network has become a hot topic of current research.Neurons and synapses are the most important parts of neural networks.In this thesis,we focus on the neuron model and synapse,and design a temperature-dependent neuron model.What's more,the establishment,renewal and pruning of synapses of temperature-dependent neuron in the neural network has been experimentally studied,which ultimately forms the completeness of the neural circuit.Firstly,based on the results and data of biological experiments,this thesis improves on the basis of Izhikevich neuron model and proposes a temperature-dependent neuron model.This model combines the traditional Izhikevich neuron model with the temperature function,and compares and analyzes the waveforms of emitting pulses of the two models,confirming that the temperature-dependent neuron model is in line with the biological basis.Secondly,this thesis applies temperature-dependent neuron model to neural network.Based on the concept of energy in Hopfield neural network,a synaptic establishment and update algorithm based on potential distribution is proposed.In view of the synaptic pruning phenomenon of microglia in biology,the synaptic pruning module is added in the network,and finally a neural circuit dominated by the stimulation points is formed,which has a great help on researching the associative memory and learning process.This thesis completed the simulation of temperature-dependent neuron model and the neural network algorithm based on potential distribution.The results show that: The temperaturedependent neuron model conforms to the results of biological experiments.That is,within the temperature threshold range,when the temperature rises appropriately,the time interval between the pulses of the neurons is shortened;when the temperature rises above the threshold,the neurons will no longer emit a pulse.In the network algorithm based on potential distribution,the process of synapse establishment and update is more purposeful than random network,so the average network and the network cost are less than random network;moreover,this algorithm designs a synaptic pruning module based on the scavenging effect of microglia in the brain on synapses,which helps to form a neural circuit dominated by the stimulation points.