摘要
针对BP算法固有的收敛速度慢和局部极小等问题，在对已有BP改进算法进行研究的基础上，提出了用于在线故障模式识别的自适应模糊加速训练算法，可在网络学习过程中，通过模糊推理，对网络拓扑结构、学习因子、惯性因子及激励函数进行自适应调整，实例验证表明该算法具有收敛速度快、不易陷入局部极小、网络综合性能好等特点。
It was found that standard BP algorithm convergesslowly and has local minimum, and some improved BP algo-rithms have their limitations. In this paper, a adaptive fuzzyaccelerating training network used in fault pattern recognitionis proposed which can adaptively adjust the sttucture of net-work, learning rate, momentum coefficient and activationfunction in the training process by fuzzy controller. The simu-lation result verifies it has better performance than normal al-gorithms.
出处
《燕山大学学报》
CAS
2000年第4期326-329,共4页
Journal of Yanshan University
基金
河北省教委科研资助!2000218
关键词
邦联模式识别
BP网络
模糊加速训练
自适应调整
fault pattern recognition, BP neural network,fuzzy accelerate training, adaptive adjusting
