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.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
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)
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
partially observable hidden state
identify the hidden state
activity decay
pre-synaptic potentials
associative memory
