Approximate Logical Dendritic Neuron Model Based on Selection Operator Improved Differential Evolution Algorithm

Abstract

In order to solve the problem that the classification effect of approximate logic dendritic neuron model is largely limited by the training effect of learning algorithm. The differential evolution algorithm of the population evolution algorithm is selected as the training algorithm of the model. Differential evolution algorithm is a branch of population evolution algorithm, which has the advantages of good robustness and easy implementation. But it also has the disadvantage that it is easy to fall into the stagnation of evolution. In order to solve this key problem, several differential evolution algorithms are investigated. Finally, we noticed that a differential evolution algorithm improved by selection operator can better solve the problem of algorithm evolution stagnation. This paper studies the training of logic dendritic neuron model using the differential evolution algorithm improved by this selection operator. In order to evaluate the performance of the algorithm training model, four representative data sets were used for experiments. When comparing the classification effect, particle swarm optimization algorithm, traditional differential evolution algorithm and genetic algorithm are selected as the comparison experiment.