Multiplication-Division Morphological Associative Memories and their Performance Analysis
- 分享到:
摘要：
形态学联想记忆是一类新颖的人工神经网络,但是它的异联想记忆效果较差.针对这个问题,基于互补的思想,提出一种形态学联想记忆方法,称为乘除形态学联想记忆.文中分析了乘除形态学联想记忆方法完全回忆记忆的条件及其噪声性质.分析和实验表明,这种方法对异联想效果显著,并具有一定的抗噪声鲁棒性,在一些情况下,能够取得更好的异联想效果.
The morphological associative memories are a kind of novel artificial neural networks,but their effects for hetero-associative memories arepoor.In order to solve this problem,based on the thought of complementary,the paper puts forward a method of morphological associative memories called multiplication-division morphological associative memories.This paper analyzed its noise property,and the condition when it can completely recall memory is deduced.The nalysis and experiment show that its effect of this method for hetero associative memories is remarkable,and has certain noise robustness.In some cases,the method can achieve better vision effect,therefore it is an effective supplement to the existing morphological associative memory methods.
作者：
冯乃勤 张亮
FENG Naiqinm ZHANG Liang(College of Information Engineering Institute of Machine Learning ＆ Big Data c. Party and government Zhengzhou University of Industrial Technology, Zhengzhou 451150,China College of Computer and Information Engineering, Henan Normal University, Xinxiang 453007,China)
机构地区：
郑州工业应用技术学院信息工程学院 郑州工业应用技术学院机器学习与大数据研究所 betway官方app 计算机与信息工程学院 郑州工业应用技术学院党政办公室
出处：
《betway官方app 学报：自然科学版》 CAS 北大核心 2016年第5期154-162,共9页
基金：
国家自然科学基金(61173071) 郑州市普通科技攻关项目(153PKJGG153)
关键词：
形态学联想记忆 异联想形态学记忆 不完全回忆记忆 互补
morphological associative memories heteroassociative morphological memories incompletely recall memory complementation
分类号：
TP18 [自动化与计算机技术—控制理论与控制工程]
