报告题目：Distribution-sharpened sampling for maximum and minimum entry localization in factorized tensors

报告人：周焕超 湘潭大学

报告时间：2026年8月3日(星期一) 10:30-11:30

报告地点：伍卓群楼三楼研讨室5

校内联系方式：邹婷婷 [email protected]

报告摘要：

  We address extremum localization in large-scale tensors stored in factorized form without full reconstruction. We propose a Distribution‑Sharpened Sampling (DSS) framework that transforms the tensor into a probability distribution, applies a monotone sharpening to boost the target entry’s dominance, and samples via sequential conditional probabilities computed by tensor contractions. A Top‑M refinement yields a robust estimate. Theoretical analysis shows sharpening improves sample efficiency and dominance, with stability under rounding errors. DSS offers a practical, theoretically sound solution for factorized high‑dimensional data.

报告人简介：

  周焕超，湘潭大学数学与计算科学学院讲师。博士毕业于东北师范大学统计专业，导师白志东教授。主要研究方向为大维随机矩阵的谱性质分析，包括极限谱分布、样本协方差矩阵的渐近行为等，致力于发展高维统计理论并将其应用于实际数据分析。发表相关论文五篇，其中包括Bernoulli、SCIENTIA SINICA Mathematica等杂志。