摘要
该文针对广泛使用的前向多层网络的ＢＰ算法存在的收敛速率低、有局部振荡的缺陷，提出了共轭梯度法改进ＢＰ算法，它在共轭梯度方向修正权值、使用概率接受原则决定目标函数值变化的取舍。同时给出了提高网络抗过配合性能的罚函数方法。实例证明：在不同的初值下，共轭梯度法均具有快的全局收敛性。
As a popular supervised training algorithm in feedforward neural networks, Back-Propagation converges slowly and immerses in vibration frequently. This paper presents a conjugate gradient momenta BP algorithm , permits the moderate increase of optimum function by probability acceptance, and gives out penalty term method to avoid overfitting.The feasibility of the method is examined with samples.
出处
《计算机工程与应用》
CSCD
北大核心
1999年第3期17-18,49,共3页
Computer Engineering and Applications
关键词
人工神经网络
BP算法
共轭梯度
Artificial Neural Network,Back-Propagation Algorithm,Conjugate Gradient
