7月3日学术报告(宫明明,美国卡耐基梅隆大学)
类别:未知 发布人:admin 浏览次数: 发布时间:2018-07-03 10:15
报告题目:   Causal and Causally-inspired Learning
报告时间:   2018年7月3日下午3:00
报告地点:   计算机学院B404
报告人:     宫明明博士
报告人国籍: 中国
报告人单位: 美国卡耐基梅隆大学计算机学院
 
报告人简介:
宫明明,美国卡耐基梅隆大学计算机学院助理教授,2017年于澳大利亚悉尼科技大学取得计算机科学博士学位,主要研究方向是 机器学习理论、因果分析、迁移学习和计算机视觉,发表方向CCF A类、B类期刊和会议近20篇,包括ICML、AAAI、IJCAI等人工智能顶会论文多篇。
 
报告摘要:
A main goal of machine learning is to discover statistical dependencies between random variables and use them to predict future observations. However, the aim of many scientific investigations is to infer how the data generating system behaves under interventions, which calls for causal structures. In this talk, I will first introduce our recent work on causal discovery from observational data. In the real world, data often have low-resolution, contain unobserved confounders, and are contaminated by measurement noise. I will show how to identify the underlying causal relations from such kind of incomplete data. Second, I will present our recent progress in addressing transfer learning/domain adaptation using causal models. The results demonstrate how causal information can be valuable for understanding distribution shift and developing more accurate predictive models.
 
邀请人: 杜博教授
上一篇:7月2日交叉学科论坛学术报告
下一篇:高性能与大数据计算培训大纲 High-Performance and Data-Intensive Computing