Venkat Chandrasekaran (Caltech) will be talking on Thursday at 10:30 am in MMCR (EE).
Title: Learning Convex Relaxations from Data
Abstract: Regularization techniques are widely employed in the solution of inverse problems in data analysis and scientific computing due to their effectiveness in addressing difficulties due to ill-posedness. In their most common manifestation, these methods take the form of penalty functions added to the objective in convex relaxation approaches for solving inverse problems. The purpose of the penalty function is to induce a desired structure in the solution, and these functions are specified based on prior domain-specific expertise. We consider the problem of learning suitable regularization functions from data in settings in which precise domain knowledge is not directly available; the objective is to identify a regularizer to promote the type of structure contained in the data. The regularizers obtained using our framework are specified as convex functions that can be computed efficiently via semidefinite programming. Our approach for learning such semidefinite regularizers combines recent techniques for rank minimization problems along with the Operator Sinkhorn procedure. (Joint work with Yong Sheng Soh)
Speaker Bio: Venkat Chandrasekaran is a Professor at Caltech in Computing and Mathematical Sciences and in Electrical Engineering. He received a Ph.D. in Electrical Engineering and Computer Science from MIT (2011), and he received undergraduate degrees in Mathematics as well as in Electrical and Computer Engineering from Rice University (2005). He was awarded the Jin-Au Kong Dissertation Prize for the best doctoral thesis in Electrical Engineering at MIT (2012), the Young Researcher Prize in Continuous Optimization (at the 4th ICCOPT of the Mathematical Optimization Society, 2013), the Sloan Research Fellowship in Mathematics (2016), and the INFORMS Optimization Society Prize for Young Researchers (2016). He is currently serving as an Associate Editor for the Annals of Statistics, the SIAM Journal on the Mathematics of Data Science, and the SIAM Journal on Applied Algebra and Geometry. His research interests lie in mathematical optimization and its interface with topics in the information sciences.