南京大学计算机软件新技术国家重点实验室 

摘 要：

Machine learning methods have demonstrated great success in many fields, but most of them are lack of interpretability and stability. Causal inference is a powerful modeling tool for explanatory analysis, which might enable current machine learning to make an explainable and stable prediction. In this talk, we will show some new challenges of estimating causal effect in the wild big data scenarios, including (1) high dimensional and noisy variables, and (2) differentiation among variables. To address these challenges, we proposed Data-Driven Variable Decomposition (D2VD) and Differentiated Confounder Balancing (DCB) algorithms. Moreover, by marrying causal inference with machine learning, we proposed a causal regularizer to recover the causations between predictors and response variable and designed a stable learning algorithm for stable prediction across unknown testing data.

报告人简介：

Kun Kuang is a Ph.D. candidate in the department of computer science at Tsinghua University. He was a visiting researcher at Graduate School of Business, Stanford University. His main research interests include data-driven causal analysis, high dimensional inference, and interpretable and stable learning. He has published several papers in top data mining and machine learning conferences/journals of the relevant field such as SIGKDD, AAAI, and ICDM, etc.

时间：5月9日  14:40-15:20

地点：计算机科学技术楼221室