An Improvement Of A Simplex Evolutionary Algorithm And Its Application In Neural Networks

Neural network is an intelligent algorithm model that simulates human thinking learning.It can solve nonlinear classification problems and related problems such as pattern recognition in practice.However,due to its own problems,it is easy to fall into the local optimal solution and the convergence speed is slow.Therefore,this thesis adopts an Improved Surface-Simplex Swarm Evolution Algorithm and uses it to optimize the structural parameters of the BP and RBF neural network.Surface-Simplex Swarm Evolution Algorithm is a new type of swarm intelligence optimization algorithm.This algorithm has a good convergence effect by establishing a surface-simplex neighborhood search operator and a multi-states search mechanism for particles.In order to further improve the performance of the algorithm,this paper improves the original algorithm's polymorphic balanced search mechanism,and proposes an improved an Improved Surface-Simplex Swarm Evolution Algorithm.This algorithm retains the advantages of the original algorithm.Perform performance testing with standard test functions and compare ISSSE with the original algorithm and other intelligent optimization algorithms.Experiments show that for most test functions,the Improved Surface-Simplex Swarm Evolution Algorithm improves the accuracy of the algorithm's solution and the convergence speed of the algorithm.The improved algorithm is introduced into the BP and the RBF neural network to optimize the structural parameters of the network,and then the optimized neural network is trained and tested using the standard data set UCI.The experiments show that the improved simplex evolution algorithm is more suitable for the BP and RBF network.The optimization of parameters has better classification effect.