活动时间：2026-04-19 10:00

活动地点：2号学院楼2432会议室

主讲人：杨建峰

主讲人中文简介：

杨建峰，男，南开大学统计与数据科学学院教授、博导，试验设计团队成员。主要从事统计试验设计和大数据分析方向的研究，在统计学顶级期刊Biometrika、Technometrics、Statistica Sinica、Science China-Mathematics等国内外高水平学术期刊发表论文近50篇；主持国家自然科学基金项目4项（青年基金1项、面上项目3项）和天津市自然科学基金面上项目1项。曾访问美国威斯康辛大学麦迪逊分校统计系和美国加州大学洛杉矶分校统计系，现为天津市现场统计研究会副理事长，中国现场统计研究会试验设计分会、中国数学会均匀设计分会、中国现场统计研究会多元统计分会常务理事，泛华统计协会终身会员，美国《数学评论》评论员。入选天津市“131”创新型人才培养工程第二层次人选、天津市教学名师后备人选；曾获教育部高等学校科学研究优秀成果奖（科学技术）自然科学二等奖、天津市教学成果奖特等奖及一等奖等省部级科研和教学奖励；荣获过南开大学“良师益友”十佳奖、南开大学感念大师系列奖项等荣誉称号；作为支部书记，曾获校级优秀共产党员和优秀党务工作者称号；带领支部获第三批“全国党建工作样板支部”，校级首批党建工作样板支部，校级先进基层党组织，校级“双带头人”教师党支部书记工作室，及校级“创最佳党日活动”一等奖等荣誉。

活动内容摘要：

Order-of-addition experiments are crucial for exploring component effects across fields, but existing linear models may fail to capture nonlinear relationships and uncertainties. In contrast, Gaussian process models excel at handling nonlinear systems and quantifying uncertainties. However, their covariance function depends on a distance metric to measure permutation dissimilarity—a limitation we address here by proposing a Gaussian process model for order-of-addition experiments based on Kendall's tau distance. We focus on constructing maximin Kendall's tau distance designs, verifying their asymptotic D-optimality under the Kendall's tau distance-based Gaussian process model, and establish an upper bound on the minimum Kendall's tau distance as the efficiency benchmark. These space-filling designs exhibit strong robustness even when the true model is unknown or misspecified. We propose multiple construction methods, some of which achieve the upper bound while others guarantee a notably large minimum Kendall's tau distance. Simulations confirm two key findings, namely that the proposed Kendall's tau-based Gaussian process model outperforms existing Gaussian process models, and the maximin design outperforms alternative methods under model-unknown scenarios.

主持人：戚良玮