4/26/18 | 4:15pm | E51-345
Reception to follow.
Abstract: A recent thrust of work, occurring at the intersection of optimization and data science, seeks to answer the following question: which statistical assumptions on input data lead to tractable nonconvex optimization problems? Most existing work on this topic applies to smooth minimization problems, which satisfy local strong convexity conditions. In contrast, relatively little work addresses nonsmooth problems, which typically lack local strong convexity. In this talk, I will discuss recent work on this topic, carried out by myself and others.
Bio: Damek Davis received his PhD in mathematics from University of California, Los Angeles in 2015. In July 2016 he joined Cornell University's School of Operations Research and Information Engineering as an Assistant Professor. In his research, Damek formulates and solves continuous optimization problems, with a focus on those which arise in machine learning and signal processing. In the past, his research has received several awards including both the NSF graduate (2010) and math postdoctoral fellowships (2015), the Pacific Journal of Mathematics Dissertation Prize (2015), and the INFORMS Optimization Society student paper prize (2014).