报告摘要: In this talk we consider the problem of graph clustering -- identifying densely connected groups of nodes in a graph. Graph clustering is an important sub-routine in various data-intensive applications including social networks and recommendation systems.We use spectral and convex optimization techniques to devise efficient clustering algorithms. Theoretical analysis are provided to show that these algorithms have strong performance under a large range of settings. We show that there is an unavoidable trade-off between the computational considerations (how much running time is needed) and the statistical considerations (how much noise can be handled). If time permits, we will discuss several extensions to the problem and algorithms, including time-varying graph and overlapping clusters.

报告人信息: Yudong Chen is currently a postdoc in the EECS department at the University of California, Berkeley in the group of Prof. Martin J. Wainwright. He obtained his Ph.D. in Electrical and Computer Engineering from the University of Texas at Austin in 2013, and his B.S. and M.S. from the Department of Automation, Tsinghua University. His research interests include machine learning, high-dimensional and robust statistics, and convex optimization, with applications in social networks, recommendation systems and air traffic control.

联系人：李力