摘要
为确定多层人工神经网络的权值和阈值建立了混合求解方法,即迭代前期采用BP算法,而迭代后期采用梯度优化法进行计算.通过BP算法、梯度优化法和混合算法对相同实例网络权值和阈值的计算,验证了所提出的混合算法的有效性,实现了权值和阈值的快速准确计算.
Mixed method was built to calculate the weights and thresholds of multi - layer neural network by utilize BP algorithm and terraced optimizations. Comparing mixed method with BP algorithm and terraced method, validity of the method was testified, which results in rapid computing of weights and thresholds.
出处
《机械与电子》
2007年第4期63-65,共3页
Machinery & Electronics
基金
国家自然科学基金资助项目(59835050)
关键词
混合算法
梯度优化法
神经网络
权值
阈值
mixed method
terraced optimizations
neural network
weights
threshold
