摘要
当混沌神经网络的输入发生较大变异时,网络混沌运动偏离了原有混沌吸引域,从而丧失了对原有被储存样本模式的记忆。本文针对神经网络的时空特征,采用混沌控制的钉扎反馈方法,使网络重新恢复记忆。通过对应用例的仿真实验表明,对神经网络时空系统的混沌控制,钉扎反馈控制是一种值得推荐的方法;神经网络的混沌控制增强了网络的容错能力及其鲁棒性,进而提高了混沌神经网络的实用性。
An associative neural network with chaotic neurons is used in the paper. The neurons of the neural nework is interconnected through a conventional auto-associative matrix of synaptic weights. This chaotic neural network is a spatiotemporal system. After the input of a chaotic neural network appears large differentiation from the reference samples, the dynamic memory of the neural network is lost. The feedback pinning can be used to control chaos of the neural network system and retriev dynamic memory. In this paper, the dynamic associative memory and retrieval for faults of broken bars in a three phase asynchronous motor is simulated by a chaotic neural network with the feedback pinning control.
出处
《模式识别与人工智能》
EI
CSCD
北大核心
2000年第4期470-473,共4页
Pattern Recognition and Artificial Intelligence
关键词
神经网络
时空混沌控制
三相异步电动机
Neural Network, Spatiotemporal Chaos Control, Dynamic Memory and Retrieval, Feedback Pinning
