具有复杂约束混料试验的渐近D-最优设计  

Asymptotically D-Optimal Design of Mixture Experiment with Complex Constraints

作　　者：李光辉[1] 张崇岐[1] 

机构地区：[1]广州大学经济与统计学院,广州510006

出　　处：《应用概率统计》2017年第2期203-220,共18页Chinese Journal of Applied Probability and Statistics

基　　金：国家自然科学基金项目(批准号:11671104)资助

摘　　要：复杂约束条件下试验设计区域极不规则,通常难以得到精确的最优设计.本文构造一种针对混料试验设计的随机搜索算法(MDRS),在具有复杂约束的区域内由Monte-Carlo方法产生一组初始点集,并通过MDRS算法迭代至逼近最优点集.通过实例验证,这种方法是有效的.它可以作为衡量其他设计的一个标准,即只有当给出的其他设计优于近似的最优解时才是有效.It is difficult to get an accurate optimum design when the experimental design area is very irregular under complex constraints. This paper constructs a random search algorithm for mixture experiments designed （MDRS）. Firstly, generating an initial points set in areas with complex constraints by the Monte-Carlo method, then use MDRS algorithm iterative to approximate optimum set. By way of example verification, this method is effective. It can be used as a standard measure of other designs, that is the only effective when given superior to other designs approximate optimal solution.

关 键 词：MONTE-CARLO方法 D-准则 信息矩阵 

分 类 号：O212.6[理学—概率论与数理统计]