Simplification Models Of Neuron Based On Cell Characteristics

In recent years,with the rise of brain science,building a large neuron network model has become one of the most important ways to simulate brain functions.Although existing neuron detailed models can accurately reflect the electrophysiological behavior of real neurons,the huge computation cost can affect the efficiency of network models.Therefore,building a simplified model of neurons is the key to solve this problem.Reduced neuron model not only simplifies the geometric morphologic complexity of neurons,but also improves the efficiency of model operation while retaining the output characteristics of neurons,which has important research significance.This thesis studies the problems faced in the process of neuron model simplification,and the main contents include:(1)At present,most of the existing neuron simplified models need to manually adjust the parameters in the simplification process,and they suffer from low accuracy and low cell generality.Therefore,this thesis proposes a simplified neuron model based on Horton policy coding.This model applies the river classification Horton strategy to the coding process and achieves the unity of different morphologic neuron coding thresholds.Then,the simplification of the detailed neuron model is realized by the steps of clustering,simplification and consolidation.Experiments show that the proposed simplified model has the good accuracy and effectively simulate the electrical responses of real neurons.(2)The simplified neuron model based on Horton strategy can not consider the influence of passive electrical characteristics such as the membrane resistance and the membrane capacitance on the simplified performance.Meanwhile,the simplified accuracy of this model is needed to improve under the partial stimulation type.For this reason,a simplified model of neuron based on cell branching characteristic coding is proposed.The model combines dendritic branching characteristics in the coding process and simplifies the geometric shape of the detailed model inspired by the idea of reserving membrane surface area.Following this,by simplifying the ratio of volume to proportion as a proportional factor,a simplified model is constructed by mapping the ion channel and the electrophysiological characteristics of neuron.Experimental results show that the proposedmodel significantly improves the simplified accuracy and accurately simulate the output characteristics of the detailed model.In summary,this thesis proposes two simplified neuron models,which simplify the morphological complexity of the neurons and preserve the output characteristics of the real neurons,leading to an effective simplification of the neuron detailed model.