摘要
基于育种算法的BP网络学习算法,用育种算法替代传统BP算法中的梯度下降法,使得改进后的算法具有不易陷入局部极小、泛化性能好等特点。该算法应用于径流预测的实验证明:这种算法能够明显减少迭代次数、提高收敛精度,其泛化性能也优于传统BP算法。
The BP neural networks learning algorithm based on Breeding Algorithm is proposed in this paper. Among this algorithm, breeding algorithm has substituted the gradient descent method in traditional BP algorithm; this new algorithm does not easily trapped local minima and has better generalization. The algorithm is applied to neural network＇s training in runoff forecasting. The results show： this algorithm can reduce number of training and error obviously; it has better generalization than traditional BP algorithm.
出处
《水力发电》
北大核心
2008年第8期12-13,42,共3页
Water Power
基金
国家自然科学基金重点项目(50539140)
国家自然科学基金项目(50679098)
关键词
BP网络
育种算法
径流预测
BP neural networks
breeding algorithm
runoff forecasting
作者简介
王羽（1975-），男，四川成都人，硕士，工程师，从事电力营销工作．
