國立政治大學統計學系
學 術 演 講
主講人：廖耿德助理教授學 術 演 講
國立清華大學統計與數據科學研究所
題 目：A Novel Framework for CATE Estimation: The Intersection of Theory
and Practice
時 間：民國114年10月13日 (星期一) 下午1：30
地 點：國立政治大學逸仙樓050101教室
摘 要：
In this talk, I will present our recent progress on conditional average treatment effect (CATE) estimation using machine learning. I will begin by introducing industrial applications of machine learning-based CATE estimation, and discussing the limitations we observed in practice. I will then present our theoretical framework able to disentangle model-wise and algorithm-wise estimation bottlenecks. Finally, I will demonstrate how this theoretical analysis can be translated into a practical model training paradigm that effectively addresses the identified limitations.
