摘要
针对"示例学习的最大复合问题及算法"一文(计算机学报,1997,20(2):139-144)中提出的FCV算法给出了一种优化改进。改进的主要思想是省略FCV算法中建立扩张矩阵,寻找公共路径这一步,直接从评价矩阵中记录选择子得到公式,从而生成规则。实验表明,优化后的算法在时间和空间性能上都有提高,其泛化能力明显高于FCV算法。
This paper gives an improved version of the FCV algorithm which has proposed in ＂the maximum complex problem in learning from examples and its greedy algorithm＂,Chinese J.Computers, 1997,20（2）： 139-144.The main idea of improvement is to remove the step of establishing extension matrix and searching common paths in FCV algorithm,and then directly generate rules with saving selectors in evaluation matrixes.Experiments show that the given improved algorithm is significantly superior to the original FCV algorithm in terms of time and space computalional complexity and generalization capability.
出处
《计算机工程与应用》
CSCD
北大核心
2008年第2期57-59,79,共4页
Computer Engineering and Applications
基金
国家自然科学基金( the National Natural Science Foundation of China under Grant No.60473045
No.60573069) 。
关键词
扩张矩阵
评价矩阵
选择子
公共元素
extension matrix
evaluate matrix
selector
common element
