多重加权多值指数双向联想记忆网络及其表决性能  

Multiple Weighted Multivalued Exponential Bidirectional Associative Memory Network and Its Voting Performance

作　　者：陈松灿[1] 蔡骏[1] 

机构地区：[1]南京航空航天大学计算机科学与工程系

出　　处：《计算机学报》2001年第2期209-212,共4页Chinese Journal of Computers

基　　金：国家自然科学基金! (6 970 10 0 4)资助

摘　　要：Wang和陈等利用各自提出的二值指数双向联想记忆模型 (e BAM)及其改进型 e BAM(Ie BAM) ,分别构造了由多个 e BAM和 Ie BAM组成的多重 e BAM(Multi- e BAM)和多重 Ie BAM(Multi- Ie BAM)的信念组合模型 ,使之可模拟多个专家的表决 .该文在此基础上 ,借助陈提出的多值 e BAM(MVe BAM) ,提出了多重多值 e BAM(Mul-ti- MVe BAM) ,对 Multi- e BAM和 Multi- Ie BAM进行了两方面的推广 :一是将二值表示推广到多值表示 ,以此可以处理现实中的多值数据 ;二是将原有模型中具有同等权威度的各专家推广到各具不同的权威度的专家 ,以此模拟更实际的表决情形 .文中借助能量函数证明了所提模型的渐近稳定性 ,以保证其实际可用 .计算机模拟证实了模型的可行性 .Wang and Chen, together with their coworkers, adopted their binary exponential bidirectional associative memory(eBAM) and improved eBAM(IeBAM) to build their multiple eBAM(Multi eBAM) and multiple IeBAM(Multi IeBAM) belief combination models respectively composed of eBAM and IeBAM to mimic the voting of many experts. In this paper, on the basis of their models and with the use of Chen multivalued eBAM(MVeBAM), a new multiple weighted MVeBAM(Multi WMVeBAM) model is presented which extends both Multi eBAM and Multi IeBAM in two aspects: one is the extension from binary to multivalued data format , the other is to apply different weights to all experts so that the proposed model can mimic a voting process in practice. By defining an energy function, the stability of the Multi WMVeBAM in synchronous and asynchronous updating modes is proven which ensures its applicability in the real world. Finally computer simulations confirm its feasiblity.

关 键 词：神经网络 表决性能 多重联想记忆网络 证据推理 信息处理 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391[自动化与计算机技术—控制科学与工程]