摘要
为提高神经网络传统BP算法的训练速度,以3层神经网络为例,通过对权值的分析与优化,推导出改良的BP算法——双权值迭代优化法,并对该算法与传统算法进行了比较,通过比较发现,新算法在保证精度的前提下可节省训练时间,同时对该算法特点进行了总结。
In order to improve the training speed of traditional BP algorithm for neural network,with three-tier neural network,through the analysis and optimization for weights,derives improved BP algorithm-double weights iterative optimization,and compares this algorithm with the traditional method,by comparison,the new algorithm significantly saves training time besides ensuring accuracy.The characteristics of this algorithm were summarized.
出处
《齐齐哈尔大学学报（自然科学版）》
2010年第4期11-15,共5页
Journal of Qiqihar University(Natural Science Edition)
关键词
神经网络
BP算法
双权值迭代优化法
neural network
BP algorithm
double weights iterative optimization
