Research On RBF Neural Network Algorithm Based On Ant Colony Algorithm

Radial basis neural network(RBFNN)has been widely used in pattern recognition,image processing,prediction,time series analysis and other fields due to its unique advantages.Ant Colony Optimization(ACO)is a kind of bionic algorithm.A large number of studies have shown that ACO,compared with other intelligent optimization algorithms,can achieve the global optimal solution.This paper briefly summarizes the basic theory of RBF neural network,the theoretical basis of development trends and ant colony algorithm,and the existing improved methods,focusing on how to apply the improved ant colony algorithm to RBF neural network to solve practical problems.Firstly,an ant colony optimization algorithm model was established based on initial pheromone distribution and secondary volatilization.Aiming at the defects of the traditional ant colony optimization algorithm,an ant colony optimization algorithm was proposed to improve the initial pheromone and establish the secondary volatilization of pheromone.The effectiveness of the ant colony optimization algorithm is tested by four sets of classical traveling salesman problems.Experimental results show that the ant colony optimization algorithm can accelerate the convergence speed of the algorithm and improve the accuracy of the algorithm.Secondly,RBF neural network model was established based on ant colony optimization algorithm.Aiming at the defects of basic RBF neural network,this paper makes use of the global search and hidden parallel ability of ant colony optimization algorithm to optimize the parameters of RBF neural network synchronously,and verifies the effectiveness of the algorithm through function fitting experiment,and obtains that the RBF neural network based on ant colony optimization algorithm has a higher degree of function fitting and faster fitting speed.Thirdly,RBF neural network model is established based on improved ant colony clustering algorithm.Taking advantage of the improved ant colony clustering algorithm,the number and position of the center of RBF neural network are determined,the weight of RLS is calculated,and the gradient descent method is used to adjust the center value,width of basis function and weight.At the same time,the Shanghai composite index prediction model based on the improved RBF neural network is established.The simulation results show that the RBF neural network based on the improved ant colony clustering algorithm has smaller prediction error and better prediction performance for the Shanghai composite index.Finally,the conclusion of the essay.Fig.21,Table 11,50 references...