Yan Liu - When Big Meets Complex: Sparse Temporal Causal Models for Learning in High-dimensional Time-Series Data
Abstract: Many emerging applications of machine learning, such as social media analysis, climate modeling, and computational biology, involve time series data with inherent structures. In this talk, I will discuss how to develop effective machine learning algorithms to uncover the complex temporal dependencies from high-dimensional time-series data. Specifically, Granger graphical models will be introduced which allows us to model sparse temporal networks from time series data by appealing Granger causality and sparse learning. I will discuss the practical challenges in analyzing time series data in computational biology, climate science, social media analysis applications and our solutions via Granger graphical models.
Bio: Yan Liu is an assistant professor in Computer ScienceDepartment at University of Southern California from 2010. Beforethat, she was a Research Staff Member at IBM Research from 2006. Shereceived her M.Sc and Ph.D. degree from Carnegie Mellon University in2004 and 2006. Her research interest includes developing scalablemachine learning and data mining algorithms with applications tosocial media analysis, computational biology, climate modeling andbusiness analytics. She has received several awards, including 2007ACM Dissertation Award Honorable Mention, best application paper awardin SDM 2007, and winner of several data mining competitions, includingKDD Cup 2007, 2008, 2009 and INFORMS data mining competition 2008. Shehas published over 50 referred articles and served as a programcommittee of SIGKDD, ICML, NIPS, CIKM, SIGIR, ICDM, AAAI, COLING,EMNLP and co-chair of workshops in KDD and ICDM.