学术报告:网络表示学习的理论分析及其应用

题目:网络表示学习的理论分析及其应用

主讲人:清华大学 唐杰

日期:2018年6月22日(星期五)

时间:下午15:30 - 16:30

地点:数据科学与计算机学院 A101

主持:王昌栋副教授

 

摘要:In this talk, I am going to quickly survey recent developed methodologies (DeepWalk, LINE, PTE, and node2vec) for network representation (embedding) learning, a new and important research topic in social network analysis. We did a theoretical analysis to show that all the aforementioned models with negative sampling can be unified into the matrix factorization framework with closed forms. We also provide the theoretical connections between skip-gram based network embedding algorithms and the theory of graph Laplacian. I will present the NetMF method as well as its approximation algorithm for computing network representation. NetMF offers significant improvements over DeepWalk and LINE (up to 38% relatively) for conventional network mining tasks. Finally, I will introduce several applications of NetMF.

 

个人介绍:清华计算机系副系主任、长聘副教授、清华-工程院知识智能联合实验室主任。CCF YOCSEF 侯任主席、CCF杰出会员、杰出演讲者。研究兴趣包括:社会网络分析、数据挖掘、机器学习和知识图谱。发表论文200余篇,引用10000多次。主持研发了研究者社会网络挖掘系统AMiner,收录1.36亿科研人员、2.31亿科技文献,吸引了220个国家/地区800多万独立IP访问。曾担任国际期刊ACM TKDD的执行主编和国际会议CIKM’16、WSDM’15的程序委员会主席;现任KDD’18大会副主席以及IEEE TKDE、ACM TIST、IEEE TBD等期刊编委编委。荣获北京市科技进步一等奖、中国人工智能学会科技进步一等奖、电子学会自然科学二等奖(均为第一完成人)以及国家自然基金委优秀青年基金、CCF青年科学家奖和英国牛顿高级学者基金奖。