Speaker: Alex Beutel, Carnegie Mellon University
Location: Warren Weaver Hall 1302
Date: February 24, 2016, 11 a.m.
Host: Subhash Khot
Can we model how fraudsters work to distinguish them from normal users? Can we predict not just which movie a person will like, but also why? How can we find when a student will become confused or where patients in a hospital system are getting infected? How can we effectively model large attributed graphs of complex interactions?
In this talk we will focus on understanding user behavior through modeling graphs. Online, users interact not just with each other in social networks, but also with the world around them – supporting politicians, watching movies, buying clothing, searching for restaurants and finding doctors. These interactions often include insightful contextual information as attributes, such as the time of the interaction and ratings or reviews about the interaction. The breadth of interactions and contextual information being stored presents a new frontier for graph modeling. We demonstrate that by modeling how fraudsters work, we can more effectively detect them, catching previously undetected fraud on Facebook, Twitter, Tencent Weibo and Flipkart. We can predictwhy you will like a particular movie, giving an interpretable recommendation system that has state-of-the-art accuracy with a 4-times smaller model. Last, we will discuss how to scale modeling of large hypergraphs, offering machine learning systems that scale to hundreds of gigabytes of data, billions of parameters and are up to 190-times faster than competitors. We will conclude with my vision for the future of modeling graphs, covering exciting new applications, novel modeling approaches and upcoming challenges in scalable machine learning.
Alex Beutel is a Ph.D. candidate at Carnegie Mellon University in the Computer Science Department. He previously received his B.S. from Duke University in Computer Science and Physics. His primary interest is in modeling large graphs, with his Ph.D. thesis focused on large-scale user behavior modeling, covering recommendation systems, fraud detection and scalable machine learning. His work has appeared in KDD, WWW, ICDM, SDM, AISTATS, and TKDD; and he has given tutorials on graph-based user behavior modeling at KDD, one of the premier data mining conferences, and CCS, one of the premier security conferences. He received the Best Paper Award at ACM GIS 2010, was a finalist for best paper in KDD 2014 and ASONAM 2012, and was awarded the Facebook Fellowship in 2013 and the NSF Graduate Research Fellowship in 2011. Beyond his research at CMU, Alex has worked on large-scale user behavior modeling at Facebook, Google, and Microsoft. More details can be found at http://alexbeutel.com.
Refreshments will be offered starting 15 minutes prior to the scheduled start of the talk.