CANCELLED: Towards Automating Machine Learning for Pattern Discovery in Big Attributed Networks
Abstract: Big data is often created by aggregating multiple data sources that are then modeled as big attributed networks. Examples include disease surveillance networks, the Internet of Things (IoT) networks, mobile networks, and social networks. Many applications of big data analytics are concerned with discovering complex patterns (subsets of the vertices, edges, and attributes) in an attributed network that provide interesting or unexpected information, such as the detection and/or forecasting of societal events (disasters, civil unrest), anomalous patterns (disease outbreaks, cyber-attacks), discriminative subnetworks (cancer diagnosis), knowledge patterns (new knowledge building), and storylines (intelligence analysis), among others. Applying traditional machine learning methods to real-world problems is time-consuming, resource-intensive, and challenging, however, and often requires machine learning experts to develop customized models, algorithms, and interpretation schemes for different applications.
In this talk, I will discuss our laboratory’s effort to develop a novel user-friendly machine learning system for non-experts that automates modeling, searching, and interpreting various complex patterns of interest in ubiquitous big attributed networks. We call our new approach Automatic Machine Learning for PATTern discovEry in big attRibuted Networks (AML-PATTERN). AML-PATTERN addresses the challenges unique to big attributed networks. These include (1) automatic modeling of complex patterns of interest by simultaneously modeling subgraph patterns (e.g., connected subgraphs; subgraphs isomorphic to a query graph) and attribute data; (2) automatic searches for complex patterns that are combinatorial by nature and thus cannot be addressed by traditional optimization techniques such as back propagation and stochastic gradient descent; and (3) the automatic interpretation of complex patterns based on uncertainty of network data due to different root causes, such as a lack of information and knowledge, out-of-distribution inputs, and conflicting evidence. I will present principled methods to address these fundamental challenges and demonstrate our proposed methods in several real-world applications.
Bio: Dr. Feng Chen is currently an Assistant Professor at the Department of Computer Science at the University at Albany - State University of New York, where he directs the Event and Pattern Detection Laboratory. He was previously a Postdoctoral Fellow at Carnegie Mellon University, having received his Ph.D. in Computer Science from Virginia Tech in 2012. Dr. Chen’s research interests include large-scale data mining, network mining, and machine learning, with a focus on event and pattern detection in massive, complex networks. His research has been funded by NSF, NIH, ARO, IARPA, and the U.S. Department of Transportation, and published in nearly 80 peer-reviewed journal and conference papers in data science and machine learning. Dr. Chen was awarded an NSF CAREER award in 2018 for his research on “Automatic Machine Learning for Discovering Complex Patterns in Big Attributed Networks”. He is a member of the IEEE and the ACM.