Author(s)

Qi Anzhi

Corresponding Author

Qi Anzhi

Abstract

BP is based on the steepest descent method and its application range is made. The algorithm is one of the most widely used neural network learning algorithms, which has the disadvantages of slow convergence speed and easy to fall into local minimum, which limits its application scope. When the scale reaches a certain level, the solution becomes impossible in terms of time, and the neural network is a good solution for approximate solution. In this paper, a hidden layer node estimation algorithm based on simulated annealing algorithm for single hidden layer BP neural network is proposed. The lower bound of the number of hidden nodes is determined based on experience. The number of hidden nodes is increased by simulated annealing until the end of the algorithm, and the optimal solution is obtained. The network was tested by test samples and compared with the predictions of nonlinear regression. Experiments show that the accuracy of the hidden layer nodes in the BP network hidden layer is higher and the speed is faster.

Keywords

BP Network, Simulated Annealing Algorithms, Number of Hidden Layer Nodes