正交拉丁超立方体设计(Orthogonal Latin hypercube designs, OLHDs)适用于计算机试验,是具有列正交性的一类空间填充设计。本文讨论了试验次数一般的一类正交拉丁超立方体设计在二维空间的投影均匀性,即在二维网格上的分层性质。结果...正交拉丁超立方体设计(Orthogonal Latin hypercube designs, OLHDs)适用于计算机试验,是具有列正交性的一类空间填充设计。本文讨论了试验次数一般的一类正交拉丁超立方体设计在二维空间的投影均匀性,即在二维网格上的分层性质。结果表明该设计的所有列对都可以实现在s × s网格分层;来自相同组连续不相邻的列对可以实现在s × s2和s2 × s网格上分层,某些列对还能实现在s2 × s2网格上的分层。The Orthogonal Latin hypercube designs (OLHD), which is a class of space-filling designs with column orthogonality, is suitable for computer experiments. In this paper, the projection uniformity of a class of OLHDs with more general run sizes in two dimensions is discussed, i.e., the grid layering properties. The results show that the design can achieve stratifications on s × s grids in any two dimensions;most column pairs can achieve stratifications on finer s2 × s and s × s2 grids when the two columns are from the same group that are not adjacent to each other, and some column pairs achieve stratifications on s2 × s2 grids.展开更多
文摘正交拉丁超立方体设计(Orthogonal Latin hypercube designs, OLHDs)适用于计算机试验,是具有列正交性的一类空间填充设计。本文讨论了试验次数一般的一类正交拉丁超立方体设计在二维空间的投影均匀性,即在二维网格上的分层性质。结果表明该设计的所有列对都可以实现在s × s网格分层;来自相同组连续不相邻的列对可以实现在s × s2和s2 × s网格上分层,某些列对还能实现在s2 × s2网格上的分层。The Orthogonal Latin hypercube designs (OLHD), which is a class of space-filling designs with column orthogonality, is suitable for computer experiments. In this paper, the projection uniformity of a class of OLHDs with more general run sizes in two dimensions is discussed, i.e., the grid layering properties. The results show that the design can achieve stratifications on s × s grids in any two dimensions;most column pairs can achieve stratifications on finer s2 × s and s × s2 grids when the two columns are from the same group that are not adjacent to each other, and some column pairs achieve stratifications on s2 × s2 grids.
文摘针对海鸥优化算法(Seagull optimization algorithm,SOA)收敛速度慢、寻优精度低以及搜索能力差等缺陷,提出一种融合自适应权重与Levy飞行的拉丁超立方体海鸥优化算法(Latin Hypercube Seagull Optimization Algorithm based onadaptive Weights and Levy flight,ALLSOA)。首先使用拉丁超立方体初始化海鸥种群,使海鸥种群全空间填充,分布更加均匀;其次在海鸥迁徙阶段,添加自适应权重因子,提高算法的搜索能力,加快算法收敛速度;最后在海鸥觅食阶段,采用Levy飞行策略,增加算法的多样性与跳出局部最优的能力,提高寻优精度。本文采用23个基准测试函数对改进算法进行测试,并利用图像分割来检验算法的有效性。试验结果表明,ALLSOA在收敛速度、寻优能力等方面表现更优。