用于神经网络容错的动态冗余BP算法  

A DYNAMIC REDUNDANCY BP ALGORITHM APPLIED IN THE FAULT TOLERANCE OF NEURAL NETWORKS

作　　者：许荔秦[1] 胡东成[1] 

机构地区：[1]清华大学自动化系,北京100084

出　　处：《电子与信息学报》2001年第10期984-988,共5页Journal of Electronics & Information Technology

基　　金：国家自然科学基金;清华大学博士学位论文基金

摘　　要：多层感知器(MLP)的容错性传统上采用改进算法和部件冗余方法。该文提出了一种动态冗余BP算法,这种方法在传统的带冲量项的自适应BP算法的学习过程中,根据各权值“重要度”的不同选取重要的权值进行冗余处理。该算法能有效地提高网络的容错能力,与学习中注入故障这一典型的容错改进算法相比,尽管容错能力并不突出,但相对可节省大量的学习时间。There are two main types of approaches in the research of fault tolerance of Multilayer Perceptrons(MLP): improvement in the learning algorithm and component redundancy after training. A dynamic redundancy BP algorithm is presented. In the training steps of the conventional adaptive BP algorithm with a momentum term, the most important weights are replicated based on their 'significance'. Applying the algorithm the fault tolerance of a network can be improved effectively. Compared with fault injection while training - a typical improved learning algorithm, although this dynamic redundancy algorithm gives no prominence in fault tolerance, the training time can be greatly reduced.

关 键 词：多层前向神经网络 容错 冗余 BP算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]