摘要
在讨论协同学习算法和广义逆关系的基础上，指出了最小二乘广义逆的求解算法都可以看作是协同学习算法，从而大大丰富了协同学习算法的种类．理论和实验表明，Ｈａｋｅｎ提及的学习算法在某些性能指标上不如通常的Ｍｏｏｓｅ－Ｐｅｎｒｏｓｅ广义逆求解算法，因此，在具体应用场合，可考虑使用通常的Ｍｏｏｓｅ－Ｐｅｎｒｏｓｅ广义逆求解算法进行协同学习．
Based on the relation between synergetic learning algorithms and generalized inverses,all the algorithms for computing least square generalized inverses can be considered as synergetic learning algorithms,which enriches the variety of synergetic learning algorithms considerably.Experiments show that some performance indexes of Haken's learning algorithms are worse than those of standard algorithms for computing MoosePenrose generalized inverses.So,those standard algorithms rather than Haken's learning algorithms are considered to be used for synergetic learning in some appilication situations.
出处
《上海交通大学学报》
EI
CAS
CSCD
北大核心
1998年第8期26-30,共5页
Journal of Shanghai Jiaotong University
基金
国防预研基金
国家自然科学基金
关键词
协同学
协同学习算法
广义逆
模式识别
synergetics
synergetic learning algorithm
generalized inverse
least square generalized inverse
MoosePenrose generalized inverse
