时 间：2023年6月9日（周五）15：30-16：30
地 点： 理科大楼A1114室
报告人：邹长亮 南开大学教授
主持人：唐炎林 研究员
摘 要：
Conformal inference is a popular tool for constructing prediction intervals (PI). We consider here the scenario of post-selection/selective conformal inference, that is PIs are reported only for individuals selected from an unlabeled test data. To account for multiplicity, we develop a general split conformal framework to construct selective PIs with the false coverage-statement rate (FCR) control. We first investigate the Benjamini & Yekutieli (2005)’s FCR-adjusted method in the present setting, and show that it is able to achieve FCR control but yields uniformly inflated prediction intervals. We propose a novel solution to the problem, named as selective conditional conformal predictions, which entails performing selection procedures on both calibration set and test set and construct marginal conformal prediction intervals on the selected sets by the aid of conditional empirical distribution obtained by the calibration set. Under a unified framework and exchangeable assumptions, we show that our proposed method can exactly control the FCR. More importantly, we provide non-asymptotic miscoverage bounds for a general class of selection procedures beyond exchangeablity and discuss the conditions under which our method is able to control the FCR. As special cases, the proposed procedure with quantile-based selection or conformal p-values-based multiple testing procedures enjoys valid coverage guarantee under mild conditions. Numerical results confirm the effectiveness and robustness of our method in FCR control and show that it achieves more narrowed PIs over existing methods in many settings.
报告人简介：
南开大学统计与数据科学学院教授。2008年于南开大学获博士学位，随后留校任教。主要从事统计学及其与数据科学领域的交叉研究和实际应用。研究兴趣包括：高维数据统计推断、大规模数据流分析、变点和异常点检测等，在Ann.Stat.、Biometrika、J.Am.Stat.Asso.、Math. Program.、Technometrics、IISE Tran.等统计学和工业工程领域期刊上发表论文几十篇，主持国家自然科学基金委重点项目、重大项目课题等。