Abstract: Aiming at the problem that the performance of harmony search algorithm is influenced by its parameters while optimizing BP network, an improved harmony search algorithm based on parameters dynamic changes with BtW is proposed and used in BP network optimization, it can improve the convergence rate and accuracy of BP network. According to the characteristics of parameters, nonlinear function transform method decided by BtW is adopted, meanwhile PAR and BW are dynamic adjusted. Then the connection and bias weights of BP network is optimized by the improved harmony search algorithm. The experimental results show that, this method has effectively improved the performance of harmony search algorithm in BP network optimization and has enhanced the training speed and accuracy of BP network.

Key words: harmony search algorithm, BP network, Best-to-Worst（BtW）, network optimization

摘要： 针对和声搜索算法参数影响其优化BP神经网络的性能问题，提出了一种可有效提高BP神经网络收敛速度和准确度的基于BtW参数动态变化的改进和声算法，同时用于BP网络优化。算法根据和声搜索参数的特点，采用以BtW为自变量的非线性函数变换方法，对微调概率PAR和微调幅度BW进行动态调整，利用改进的和声搜索算法对BP神经网络的连接权和偏置值进行优化。实验结果表明，该算法有效改善了和声搜索算法在BP神经网络优化中的性能，提高了BP网络的训练速度和预测的准确度。

关键词: 和声搜索算法, BP神经网络, BtW, 网络优化