题目：Model-free Trustworthy Learning

报告人：夏寅 教授（复旦大学）

地点：致远楼108室

时间：2025年10月20日 10:00-11:00

Abstract：In this talk, we present new developments in uncertainty quantification that aim to enhance statistical reliability and interpretability in complex learning scenarios. First, we introduce a Synthetics, Aggregation, and Test inversion (SAT) approach for merging diverse and potentially dependent uncertainty sets into a single unified set. The procedure is data-light, relying only on initial sets and their nominal levels, and it flexibly adapts to user-specified input sets with possibly varying coverage guarantees. To address this, SAT constructs and aggregates novel synthetic test statistics, and then derive merged sets through test inversion. Our method leverages the duality between set estimation and hypothesis testing, ensuring reliable coverage in dependent scenarios. A key theoretical contribution is a rigorous analysis of SAT's properties, including its admissibility in the context of deterministic set merging. We then turn to the challenge of controlling misclassification rate in large-scale multi-class classification. We propose a unified PSP (Pre-classification, Selective p-value construction, and Post-classification) framework that controls general group-wise error rates. PSP is distribution-free and offers valid finite-sample guarantees. We further establish general power optimality theories for PSP from both non-asymptotic and asymptotic perspectives. Together, these frameworks offer principled solutions for trustworthy learning with rigorous error control.

报告人简介: 夏寅，复旦大学管理学院统计与数据科学系教授，博导，博士毕业于宾夕法尼亚大学，曾在美国北卡大学教堂山分校任tenure track助理教授。她2016年入选中组部千人计划青年项目；2020年获得国家自科基金优秀青年基金资助。夏寅目前担任AOS和AOAS的副主编，她的研究方向包括高维统计推断、大范围检验及应用等，她在JASA, AOS, JRSSB, Biometrika等期刊上发表三十余篇论文。

欢迎各位参加！