摘要
本文给出了一种用循环前馈神经网络来表示具有ｎ个状态的离散马尔科夫模型，及其二次能量函数，并根据梯度下降规则，给出了相应的权值修改规则。利用该神经网络方法，可以有效地对容错硬件进行可靠性设计。
This paper gives a feed forward recursive neural network to represent the discrete time n state Markov model.A quadratic energy function for the neural network is presented and appropriate update rules for the weights are derived using the least mean square gradient descent technique.This method can be used to efficiently design the reliability of fault tolerant hardware.
出处
《计算机工程与科学》
CSCD
1999年第3期10-14,共5页
Computer Engineering & Science
