- Outlier Detection for Graph Data. Manish Gupta (Microsoft), Jing Gao (Buffalo), Charu Aggarwal (IBM), Jiawei Han (UIUC)
- Social Network Analysis: A Computational Perspective, Jaideep Srivastava (UMN)
- Big Graph Mining: Algorithms, Anomaly Detection, and Applications. U Kang (KAIST), Leman Akoglu (Stony Brook), Duen Horng (Polo) Chau (Georgia Tech)
Title: Outlier Detection for Graph Data
Speaker: Manish Gupta (Microsoft India)
Abstract: Outlier detection has been studied in the context of many research areas like statistics, data mining, sensor networks, environmental science, distributed systems, spatio-temporal mining, etc. Outlier detection has been studied on a large variety of data types including high-dimensional data, uncertain data, stream data, graph data, time series data, spatial data, and spatio-temporal data. We present an organized picture of recent research in outlier detection for graph data for both static as well as dynamic graphs. We begin by motivating the importance of graph outlier detection and briefing the challenges beyond usual outlier detection. Static graph outlier detection techniques include Minimum Description Length techniques, techniques based on egonet metrics and random field models. For dynamic graphs, we discuss graph similarity based algorithms, evolutionary community based algorithms and online graph outlier detection algorithms. We also present applications where such techniques have been applied to discover interesting outliers.
Title: Social and Behavioral Analytics (SBA): Mining Behaviors of a Socially Connected World
Speaker: Jaideep Srivastava (Computer Science and Engineering, University of Minnesota)
Abstract: That computing is pervasive in society today is almost a cliché. However, treating this cliché as one misses an essential truth of our times, namely, that our social existence has become intimately intertwined with computing and communication technology. Smartphones and broadband communication have fundamentally changed the way we interact and socialize, the way we seek advice, basically the way we live our lives. This ubiquitous, ‘on-the-go’ availability of both significant computing power, and high speed connectivity, has led the software industry to create all sorts of ‘apps’ which people can take advantage of – and that too almost always for free. Of course, what is hidden in plain sight is that fact that these apps collect data about people, of all kinds, in all sorts of ways, and all moments and locations; and users of these apps are usually not even aware of this. This has created a whole new world of data, at a scale, and of a type that is unprecedented in history. This opens up all sorts of opportunities to achieve things never before dreamed possible, and at the same time makes us once in a while think where we might be headed, and if that is indeed the destination we want to get to. Technology, however, is ethics agnostic, and moves at its own pace and in its own direction. The goal of this course is to discuss, in a tutorial manner, through case studies, and through discussion, what these technologies are, and where they seem to be taking us.
Title: Big Graph Mining: Algorithms, Anomaly Detection, and Applications
Speaker: U Kang (KAIST), Leman Akoglu (Stony Brook), Polo Chau (Georgia Tech)
Abstract: Graphs are everywhere: social networks, the World Wide Web, biological networks, and many more. The sizes of graphs are growing at unprecedented rate, spanning millions and billions of nodes and edges. What are the patterns and anomalies in such massive graphs? How to design scalable algorithms to find them? And what kind of real-world problems can we solve with such tools? These are exactly the goals of this tutorial.
We start with important graph algorithms central to graph mining and pattern discoveries, and describe how we can implement their fast, scalable versions using a unified framework on top of Hadoop. Then we describe graph-based anomaly detection techniques (complement of pattern discoveries) and how to scale them to massive graphs. Finally, we discuss how our aforementioned techniques can help solve large-scale, real-world problems that make impact to society, and to help solve challenging problems in visual analytics.