确定部分观测隐状态的k步记忆模型研究  

Research on k-step memory model inference of partially observable hidden state

作　　者：王作为[1] 梁晓丹[1] 张汝波[2] 

机构地区：[1]天津工业大学计算机科学与软件学院,天津300387 [2]大连民族学院机电信息工程学院,辽宁大连116600

出　　处：《华中科技大学学报（自然科学版）》2013年第S1期356-359,共4页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家高技术研究发展计划资助项目(2009AA04Z215);国家自然科学基金资助项目(60975071;61173032)

摘　　要：提出一种顺序联想记忆网络模型以解决隐状态的定位问题.该模型利用活性衰退、突触势能以及同步激活实现k步记忆,根据k步顺序历史确定隐状态.该模型是一种分布式计算模型,并行机制使得隐状态的定位不会因为存储知识的增多而效率下降,具有真正的在线计算能力.最后对记忆的迭代算法进行改进,实验结果表明改进后的模型具有更好的容错能力.A sequence associative memory network was proposed in this thesis to resolve the hidden state problem.This model utilized the cell activity decay,pre-synaptic potentials and cell-synfire mechanisms to realize the k-step memory,identify the hidden state according to k-step sequence history.This model was a distributed computing model.The efficiency of identifing the hidden state didn′t drop along with the increase of knowledge storage by use of its parallel mechanism.It possessed the real on-line computing capability.Finally,the improved memory iteration algorithm was presented. The experiment results show that the improved algorithm has better fault-tolerant ability.

关 键 词：部分观测隐状态 确定隐状态 活性衰减 前突触势能 联想记忆 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]