一、 青年教师工作论文报告

二、特邀专家报告

题目：Personalized Variable selection

摘 要：We study a new important problem of deciding whether to measure certain predictor variables for a given individual. In many real-world applications, measuring predictors can be expensive or potentially harmful. It is therefore important to only measure these predictors when they provide valuable information. Traditional variable selection methods select the variables that are most useful on average. However, for complicated models, it is common for a predictor to provide different amounts of information for different subjects or individuals, meaning that the predictor may be very useful for some individuals and useless for others. In the personalized variable selection paradigm, we consider the problem of using a fitted model to make predictions for a new observation where we have not yet measured all these costly variables. We assess the predictive value of the potentially useful predictor variables for this new observation, in order to decide which predictors are worth measuring for this observation. We introduce a novel metric called the Expected Loss Improvement Estimate (ELIE), which quantity the expected gain in predictive accuracy from measuring a missing variable. The core idea of our method is to impute a large number of plausible values for the missing predictor, then assess how much the resulting predictions change for different imputed values.

报告人：Gu Hong

专家简介：Gu Hong，加拿大达尔豪斯大学统计系教授、博士生导师。1999年获香港大学博士学位。主要研究领域为计算生物学与生物信息学，尤其专注于统计建模、推断方法的开发及其在分子进化与宏基因组学中的应用与拓展。同时，Gu Hong教授对数据挖掘方法及其应用亦有浓厚兴趣，特别是在医学数据分析方面开展了深入研究。其相关成果已发表于Journal of the American Statistical Association、Statistics in Medicine 等国际权威期刊，并主持多项加拿大自然科学与工程研究理事会（NSERC）资助课题。