Empirical Process and Modern Statistical Decision Theory

经验过程与现代统计决策理论

• 批准号: 1534545
• 负责人: Huibin Zhou
• 金额: $ 2.1万
• 依托单位: Yale University
• 依托单位国家: 美国
• 项目类别: Standard Grant
• 财政年份: 2015
• 资助国家: 美国
• 起止时间: 2015-05-01 至 2016-04-30
• 项目状态: 已结题
• 来源: https://www.nsf.gov/awardsearch/showAward?AWD_ID=1534545&HistoricalAwards=false
• 关键词: Empirical; Process; Modern; Statistical; Decision

The workshop titled "Empirical Process and Modern Statistical Decision Theory" will be held at Yale University on May 7-9, 2015. The era of big data is fundamentally changing every aspect of our life through discoveries in science, medicine, and engineering. Many innovative and intuitively appealing methodologies have been proposed to make novel and significant discoveries by analysis of complex and big data. It is timely and critically important for statisticians to develop deep, broad, and formal statistical theory to understand and justify why and when certain methodologies would work or not work, to guide statistical practice to make valid and influential contributions to our society. This workshop will bring together some of authorities in statistical decision theory and empirical Process to review the most important and influential advances in the past, to report their most recent exciting research, and to discuss their view of future developments. The workshop will provide a venue for promising young researchers to interact with these leaders of the field and each other, leading to future collaborations and discoveries. A potential outcome of this workshop will provide a guidance to researchers and professors on how and what to teach in empirical process to train our students in understanding and developing modern statistical decision theory.Empirical process has been playing a key role in developing modern statistical decision theory for a wide range of important models and significant methodologies, such as providing a unified framework for many high dimensional linear models by Gaussian width, establishing asymptotic equivalence theory of Le Cam for various statistical models by the KMT construction, and justifying the effectiveness of important algorithms in machine learning including SVM by VC dimension. In the last ten to fifteen years, the statistics research has witnessed a tremendous successes of empirical process theory in Bayesian nonparametrics, shape constrained estimation, robust estimation, minimax regret, lasso, and sparse principal component analysis. This workshop will bring together some of authorities in statistical decision theory including Lawrence Brown and Iain Johnstone, and in empirical processes including Richard Dudley and Evarist Gine, as well as some of the the most prominent researchers in high dimensional estimation, Bayesian nonparametrics, robust estimation, machine learning, and shape constrained estimation, to celebrate the most significant advances in the past and to discuss exciting future developments.

题为“经验过程和现代统计决策理论”的研讨会将于2015年5月7日至9日在耶鲁大学举行。大数据时代通过科学、医学和工程方面的发现从根本上改变了我们生活的方方面面。已经提出了许多创新和直观吸引人的方法，通过分析复杂和大数据来做出新颖和重大的发现。对于统计学家来说，发展深入，广泛和正式的统计理论来理解和证明为什么以及何时某些方法会起作用或不起作用，以指导统计实践为我们的社会做出有效和有影响力的贡献是及时和至关重要的。本次研讨会将汇集统计决策理论和经验过程的一些权威，回顾过去最重要和最有影响力的进展，报告他们最近令人兴奋的研究，并讨论他们对未来发展的看法。该研讨会将为有前途的年轻研究人员提供一个场所，与该领域的领导者进行互动，并相互交流，从而促进未来的合作和发现。本次研讨会的一个潜在成果将为研究人员和教授提供指导，指导他们如何以及如何教授经验过程，以培养我们的学生理解和发展现代统计决策理论。经验过程在发展现代统计决策理论中发挥着关键作用，适用于广泛的重要模型和重要方法，如用高斯宽度为许多高维线性模型提供了一个统一的框架，用KMT构造建立了各种统计模型的LeCam渐近等价理论，用VC维来证明支持向量机等机器学习中重要算法的有效性。在过去的10 ~ 15年里，统计学研究中的经验过程理论在贝叶斯非参数估计、形状约束估计、稳健估计、极小极大后悔、套索和稀疏主成分分析等方面取得了巨大的成功。本次研讨会将汇集统计决策理论的一些权威，包括Lawrence Brown和Iain Johnstone，以及经验过程中的一些权威，包括Richard达德利和Evarist Gine，以及高维估计，贝叶斯非参数，鲁棒估计，机器学习和形状约束估计中最杰出的研究人员，庆祝过去最重要的进步，并讨论令人兴奋的未来发展。