报告主题：Attributed Network Learning: Data, Theory, and Algorithms
报告摘要：Most machine learning algorithms assume that instances are independent of each other. It does not hold for networked data. Node representation learning (NRL) aims to learn low-dimensional vectors to represent nodes, such that actionable patterns in networks and side information can be preserved. These representations could be leveraged by off-the-shelf machine learning algorithms to conduct tasks. Data availabilities necessitate the development of NRL. Various theories and algorithms have been explored to learn from different types of networks, including large-scale, attributed, heterogeneous, dynamic, and directed networks, as well as knowledge graphs. In this talk, we focus on an increasingly-common type of networks, named attributed networks. We review the development of and recent advances in NRL for attributed networks. The main idea is to take full advantage of node attributes to remould the core components of network learning techniques, including random walks and graph neural networks.
主讲人：黄啸
时 间：2021年11月19日 下午3:00～4:00
会议ID：468 927 921
主讲人简介：
Dr. Xiao Huang is currently an Assistant Professor in Department of Computing at the Hong Kong Polytechnic University. He received Ph.D. in Computer Engineering from Texas A&M University in 2020 and B.S. in Engineering from Shanghai Jiao Tong University in 2012. He has actively published in prestigious conferences and journals including KDD, NeurIPS, AAAI, and IJCAI. His publications are well recognized by the community and have received about 1200 citations. He has given a lecture-style tutorial named “Learning from Networks” as the leading author, at KDD 2019. He is a senior program committee member of IJCAI 2021, and a PC member of WWW 2022, ICLR 2022, AAAI 2021-2022, NeurIPS 2021, ICML 2021, KDD 2019-2021, etc. He is a reviewer of many international journals. Before joining PolyU, he worked as a research intern at Microsoft Research and Baidu USA.