摘要
利用优化算法中的信赖域法,选取了合适的计算模型,并具有全局收敛性、收敛快速性。BP网络学习算法对异或问题、逼近问题进行的具有统计意义的仿真,表明了该算法的有效性及对样本数量变化不敏感的特性。
:Presents a new learning algorithm developed using a trust region method for BP networks with overall convergence and enhanced convergence rate and concludes from simulation results that the new algorithm is effective.
出处
《哈尔滨建筑大学学报》
北大核心
2001年第1期94-96,共3页
Journal of Harbin University of Civil Engineering and Architecture
关键词
BP网络
前向神经网络
信赖域法
ABPM算法
BP network; Least square method with damper;feedforward neural networks;error back propagation;trust region method;approach problem;ABPM algorithm