Roles in Networks - Foundations, Methods and Applications
Networks (or graphs) are widely used in representing complex relationships among entities. Nodes may function as the center, peripheral, clique or bridge. Role analytics is an important task in graph mining and can shed light on a variety of applications in networks. Thus, it has been studied by social scientists, database researchers and data mining community.
This tutorial aims to give a comprehensive discussion of role analytics in networks including role discovery, role-oriented network embedding and role-based applications. We first briefly revisit the basic concepts of equivalence relations which lay the foundation for later research on role analytics. Then we discuss the categorization of existing approaches for role discovery and overview some representative, distinct, and popular approaches for each category. Moreover, we focus on role-oriented network embedding approaches and propose a two-level categorization to better classify existing methods. Finally, we point out the challenges and future directions.
The goal of this tutorial is to offer a comprehensive presentation of role analytics in networks with focuses on role discovery, role-oriented network embedding and role-based applications. Specifically, it includes basic concepts of structural roles, representative approaches from both sociology and computer science, role-oriented network embedding approaches, and challenges and future directions. Role analytics is an exciting research topic studied by social scientists, database researchers and data mining community. This tutorial focuses on presenting state-of-the-art computational approaches for role analytics showing the connections across data mining, database and social science.
ICDM - Roles Analytics in Networks (Prelimary version)
- Introduction [15 min]
- Overview of role analytics in networks
- Relation between roles and communities
- Preliminaries [20 min]
- Basic concepts of role equivalences
- Different taxonomies of role analytic methods
- Taxonomy of Role Analytic Methods [25min]
- Equivalence-based methods
- Similarity-based methods
- Blockmodel-based methods
- Feature-based methods
- Embedding-based methods
- Role-oriented Embedding Methods [30 min]
- Low-rank matrix factorization methods
- Random walk based methods
- Deep learning methods
- Role-based Applications [25 min]
- Social role detection
- Role-enhanced Community Detection
- Network dynamics analysis
- Role-oriented Graph Neural Networks
- Challenges and Future Directions [20 min]
- Interpretable roles analytics
- Role analytics in different types of networks
- Evaluation on role discovery
- Joint role and community Detection
- Role representation in non-Euclidean space
- Construction of larger-scale benchmarks
Asst. Prof. Yulong Pei is an assistant professor with Department of Mathematics and Computer Science, Eindhoven University of Technology (TU/e). He received his Ph.D. (cum laude) in Computer Science from TU/e in February 2020. His research interests cover graph mining, network embedding, and text mining. He has over 30 publications including papers in top conferences, such as ICML, AAAI, CIKM, and COLING and IJCAI, and journals, such as DMKD and TKDE. He has served as the PC member of top-tier conferences including NeurIPS, ICML, AAAI, IJCAI and ECMLPKDD, and the regular reviewer for prestigious journals like TKDE and TKDD.
Dr. Pengfei Jiao is a lecture with the Center of Biosafety Research and Strategy of Tianjin University. He received the Ph.D. degree in computer science from Tianjin University, Tianjin, China, in 2018. His current research interests include complex network analysis, data mining and graph neural network, and currently working on temporal community detection, link prediction, network embedding, recommender systems and applications of statistical network model. He has published more than 50 international journals and conference papers.
Dr. Akrati Saxena is a Research Fellow at the Department of Mathematics and Computer Science, Eindhoven University of Technology (TU/e), Netherlands. She received her Ph.D. in Computer Science and Engineering from IIT Ropar, India. Her research interests cover social network analysis, complex networks, computational social science, and Fairness. She has worked extensively at Network Science and Data Science and has published at well known conferences and journals related to Network Science including ASONAM, Complex Networks, SNAM, JCSS.
Mr. Xuan Guo received the Bachelor degree in Computer Science from Tianjin University. He is pursuing a doctoral degree at the College of Intelligence and Computing, Tianjin University, P. R. China. His current research interests include complex network analysis, role discovery, network representation learning and percolation model.
Prof. George Fletcher is full professor with Department of Mathematics and Computer Science, Eindhoven University of Technology (TU/e), where he chairs the Databases group. He studies data-intensive systems, with a focus on graph analytics. He recently co-organized GRADES-NDA 2019 workshop at SIGMOD and an NII Shonan Seminar on Graph database systems 2018. He also chaired the EDBT Summer School on Graph Data Management 2015. He serves on the major PC of data management and AI conferences.
Prof. Mykola Pechenizkiy is full professor with the Department of Mathematics and Computer Science, Eindhoven University of Technology (TU/e), where he chairs the Data Mining group. In the past he co-developed and delivered a series of tutorial on handling concept drift in machine learning, including tutorials at ECMLPKDD 2012 and ECMLPKDD 2010. It was later supported with the ACM CS survey paper. He co-authored over 100 papers in data mining and its applications including publications in DMKD, TKDE, KAIS, AAAI, IJCAI, SDM and ECMLPKDD among others. He is an active contributor to the ECMLPKDD community, serving as area chair in both Research and Applied Data Science Tracks, and serving on the editorial board of DMKD.
Tutorial@ICDM2021: Roles in Networks - Foundations, Methods and Applications