摘要
对两种改进BP(Back-Propagation)算法神经网络的滤波进行了比较,通过反复的训练得出两种改进算法在权值调整和误差收敛过程中所遇到的共同点和差别,并得出初始权值、激励函数和神经元个数等因素的选取对神经网络学习的影响。
Two improved BP(Back-Propagation)algorithms are compared,which are used in ANN(Artificial Neural Network)filter.By repeating training,their common ground and the difference in the process of weight adjusting and error constringency are drawn,and the effect of neuron num-ber,exciting function and initial weight on the training of ANN is summarized.
出处
《电力自动化设备》
EI
CSCD
北大核心
2003年第6期35-36,40,共3页
Electric Power Automation Equipment
关键词
BP算法
神经网络
滤波
BP algorithms
neural network
filter