极强的可靠性和容错性是神经网络的重要特性,对神经网络其它主要特性(如并行性和自学习特性)有重大影响。对于神经网络的可靠性研究十分必要。由于近年来重新掀起了神经网络的研究热潮,对于神经网络的可靠性方面的研究与设计工作亦是刚刚起步。本文旨在对已有的研究结果进行总结与归纳,力求从中得出一些重要的结论,并提供一些有价值的研究方法。本文还对这一研究领域的发展方向进行了探讨。 Extreme reliability and fault tolerance are important features of neural networks and have a significant impact on other key neural network features, such as parallelism and self-learning features. It is necessary to study the reliability of neural network. In recent years, the wave of research on neural networks has been revived. Research and design work on the reliability of neural networks has just begun. The purpose of this paper is to summarize and summarize the existing research results and try to draw some important conclusions and provide some valuable research methods. This article also explores the development direction of this research area.