Abstract: This talks is about non-Euclidean optimization, primarily (but not only) about problems whose parameters live on a Riemannian manifold. I’ll lay particular emphasis on geodesically convex optimization problems, a vast class of non-convex optimization problems that remain tractable. I’ll recall some motivating background and canonical examples, before talking about theoretical progress. Starting with our paper in 2016, iteration complexity theory for g-convex optimization has now grown substantially: I will summarize some of the recent progress (including Riemannian acceleration, saddle-point problems, Langevin methods, etc.) while also sharing key open problems.
Bio: Suvrit Sra is an Associate Professor of EECS at MIT, and a core member of the Laboratory for Information and Decision Systems (LIDS), the Institute for Data, Systems, and Society (IDSS), as well as a member of MIT-ML and Statistics groups. He obtained his PhD in Computer Science from the University of Texas at Austin. Before moving to MIT, he was a Senior Research Scientist at the Max Planck Institute for Intelligent Systems, Tübingen, Germany. His research bridges mathematical areas such as differential geometry, matrix analysis, convex analysis, probability theory, and optimization with machine learning. He founded the OPT (Optimization for Machine Learning) series of workshops, held from OPT2008–2017 at the NeurIPS (erstwhile NIPS) conference. He has co-edited a book with the same name (MIT Press, 2011). He is also a co-founder and chief scientist of macro-eyes, an OPT+ML startup with a focus on social good.