Abstract:
Based on polynomial curve-fitting theory,a Jacobi orthogonal basis feed-forward neural network is constructed.The model adopts a three-layer structure,where the hidden-layer neurons are activated by Jacobi orthogonal polynomial functions.In view of the network,based on a weights-direct-determination method,this paper proposes a quick-determination algorithm for the number of hidden-layer neurons.First the concepts of orthogonal basis functions and Jacobi polynomial,and the basic principles of BP neural network are introduced.Then the network hidden layers is designed,the number of hidden neurons is determined,the layer connection weights is set,and the improved algorithm steps are given.Finally,comparing with the traditional matrix iterative method and Levenberg-Marquardt training algorithm,the proposed algorithm is validated.The simulation results show that the algorithm is more efficient and effective than conventional BP iterative-training algorithms.