混合故障模型下离散 Hopfield 联想存储器的鲁棒性  

Robustness of discrete Hopfield associative memories under mixed fault model

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：《清华大学学报（自然科学版）》1999年第7期46-49,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家自然科学基金;国家教委博士点基金

摘　　要：离散 Ｈｏｐｆｉｅｌｄ 联想存储器（ Ｄ Ｈ Ａ Ｍ）在联想记忆方面有很大的优点，但是在硬件实现过程中存在着各种故障。对于软硬故障同时存在情况下 Ｄ Ｈ Ａ Ｍ 的性能进行分析是很有必要的。该文对 Ｄ Ｈ Ａ Ｍ 建立了同时包含权值误差、权值断路、神经元输出固值等各种软、硬故障的故障模型，定义了概率意义下 Ｄ Ｈ Ａ Ｍ 的鲁棒性。运用概论统计的方法推导出了运行于同步方式下 Ｄ Ｈ Ａ Ｍ 的鲁棒性，推导中对结果进行了近似处理。对 Ｄ Ｈ Ａ Ｍ 系统进行了仿真实验，与利用这种鲁棒性分析方法计算进行了对比，结果相差非常小，证实了此方法的正确性，可以用于 Ｄ Ｈ Ａ Ｍ 的可靠性计算。Discrete Hopfield associative memories (DHAM) has great advantage in associative recalling. But there are kinds of faults in the DHAM implemented with hardware. The analysis on DHAM's properties under these soft and hard faults is necessary. A DHAM's mixed fault model was established, including both soft faults such as weights errors and hard faults such as weights broken connections, neuron stuck at faults. And the robustness of DHAM was derived using probability and statistic method, with some approximate reasoning using. The results obtained in the simulation on a DHAM system under this mixed fault model show very small diversity with the calculating results using this robustness analysis method. This shows that this method is right and it can be used in DHAM reliability calculation.

关 键 词：联想存储器 神经网络 鲁棒性 DHAM 混合故障模型 

分 类 号：TP333[自动化与计算机技术—计算机系统结构]