Dynamic processes, such as rumor spreading in social networks, occurrence of crimes in a city, migration of birds across continents, generate a large volume of high dimensional “asynchronous” and “interdependent” temporally and spatially stamped event data.  This type of event data is rather different from traditional iid. data and discrete-time temporal data, which calls for new models and scalable algorithms for analyzing, learning and utilizing them. In this talk, I will present a framework based on multivariate point processes, high dimensional sparse recovery, and randomized algorithms for addressing a sequence of problems arising from this context.  As a concrete example, I will also present experimental results on learning and optimizing information diffusion in web logs, including estimating hidden diffusion networks and influence maximization with the learned networks. With both careful model and algorithm design, the framework is able to handle millions of events and millions of networked entities, and achieve the state-of-the-art results.

Dr. Le Song is an assistant professor in the Department of Computational Science and Engineering, College of Computing, Georgia Institute of Technology. His principal research interests lie in machine learning methodology, such kernel methods, probabilistic graphical models, temporal data and network analysis, and the applications of machine learning to interdisciplinary problems, such as computational biology and materials science. Le Song received his Ph.D. in Computer Science from University of Sydney in 2008, and then conducted his post-doctoral research in the School of Computer Science, Carnegie Mellon University, between 2008 and 2011. Before he joined Georgia Institute of Technology, he worked briefly as a research scientist at Google.