Scholarlink

<p>Sliced space-filling designs are effective for computer experiments with both qualitative and quantitative variables, maintaining space-filling properties in the full design of quantitative variables and within each subdesign partitioned by qualitative variables. However, non-uniform designs like generalized Latin hypercube designs can offer superior predictive accuracy. Inspired by this, we propose a generalized sliced space-filling design based on a hybrid weighted energy distance criterion. One-shot and sequential construction algorithms are developed. Numerical studies show that the proposed design provides greater flexibility in run size and experimental region, while achieving better predictive performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Generalized sliced space-filling design

  • 期刊论文

  • 发表日期：

    2026年5月13日

查看全文
  • Yuning Huang,
  • Zikang Xiong,
  • Hong Qin

摘要

Sliced space-filling designs are effective for computer experiments with both qualitative and quantitative variables, maintaining space-filling properties in the full design of quantitative variables and within each subdesign partitioned by qualitative variables. However, non-uniform designs like generalized Latin hypercube designs can offer superior predictive accuracy. Inspired by this, we propose a generalized sliced space-filling design based on a hybrid weighted energy distance criterion. One-shot and sequential construction algorithms are developed. Numerical studies show that the proposed design provides greater flexibility in run size and experimental region, while achieving better predictive performance.