摘要
本文在计算机试验的基础上,提出了最小相关准则和最小距离离差准则,并将信息论中的Hamming距离和Lee距离引入到计算机试验中,证明了均匀设计在Hamming距离下的最优性和部分好格子点均匀设计在Lee距离下的最优性.基于偏差的考虑,给出了一类新的好格子点均匀设计和一个学习算法,利用这个学习算法,给出了基于Lee距离的最小距离离差准则的均匀设计表的构造方法.通过与已有的好格子点均匀设计和循环拉丁方均匀设计作比较,证明了文中的均匀设计在距离和偏差意义下有更好的均匀性.
Based on computer experiments.the concepts of minimum correlation and minimum distance aberration (MDA) are proposed.By applying the Hamming distance and Lee distance of information theory,it is shown that uniform designs are optimal with Hamming distance and some uniform designs based on good lattice point sets (UDGLP) are optimal with Lee distance.A new class of UDGLP and a learning algorithm are suggested for discrepancy.With this algorithm,a construction method for uniform designs is given based on the MDA criterion of Lee distance.Compared with other UDGLP and UDCLS,the new uniform designs have better uniformality under distance and discrepancy.
出处
《高校应用数学学报(A辑)》
CSCD
北大核心
1998年第2期167-174,共8页
Applied Mathematics A Journal of Chinese Universities(Ser.A)
基金
国家自然科学基金