小世界体系的多对多核联想记忆模型及其应用  

Small World Structure Inspired Many to Many Kernel Associative Memory Models and Their Application

作　　者：陈蕾[1,2] 陈松灿[1] 张道强[1,3] 

机构地区：[1]南京航空航天大学计算机科学与工程系 [2]南京邮电大学计算机科学与技术系,江苏南京210003 [3]计算机软件新技术国家重点实验室(南京大学),江苏南京210093

出　　处：《软件学报》2006年第2期223-231,共9页Journal of Software

基　　金：国家自然科学基金;江苏省自然科学基金;江苏省博士后科研资助计划~~

摘　　要：运用机器学习中新颖的核方法和社会网络中广泛存在的小世界现象,对Hattori等人提出的多模块多对多联想记忆模型(multi-moduleassociativememoryformany-to-manyassociations,简称(MMA)2)进行了改进,构建出了一个基于小世界体系的多对多核联想记忆模型框架(smallworldstructureinspiredmanytomanykernelassociativememorymodels,简称SWSI-M2KAMs).该框架不仅克服了原模型不能联机提交训练样本且迭代次数过多的缺陷,而且拓展了原模型的智能信息处理范围.更重要的是,通过核函数的选取,该模型框架可以衍生出更多新的多对多联想记忆模型,而且,由于小世界结构的引入,在一定程度上简化了模型的结构复杂度.最后的计算机模拟,证实了新的模型具有良好的多对多联想记忆功能.Kernel method is an effective and popular trick in machine learning, and small world network is a common phenomenon which exists widely in social fields. In this paper, by introducing them into Hattori et al＇s multi-module associative memory for many-to-many associations （（MMA）^2）, a unified framework of small world structure inspired many-to-many kernel associative memory models （SWSI-M2KAMs） is proposed. The SWSI-M2KAMs not only can store patterns online without more iteration steps, but also extend the range of the processed intelligent information. More importantly, the SWSI-M2KAMs framework can develop more new many-to-many associative memory models by selecting different kernel functions and reduce models＇ configuration complexity by using the sparse small world architecture. Finally, computer simulations demonstr ate that the constructed models have good performance on many-to-many associative memory.

关 键 词：神经网络 多对多联想记忆 核方法 小世界理论 智能信息处理 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]