Simons Workshop on Fast Iterative Methods in Optimization

Fazel was a co-organizer of the Simons Workshop on Fast Iterative Methods in Optimization, held on Oct 2-6, 2017 at the Simons Institute, UC Berkeley. This workshop was part of the semester-long program on Bridging Continuous and Discrete Optimization held Aug-Dec 2017, and Fazel was also a long-term participant of the program, which aimed to bring together experts from these two sides of the broader optimization community.

Recently, a confluence of ideas from optimization and theoretical computer science led to breakthroughs in terms of new understanding and running time bound improvements for some classic iterative continuous optimization primitives. This workshop explored these advances as well as new directions that they have opened up. Some of the specific topics were advanced first-order methods (non-smooth optimization, regularization, preconditioning), structured optimization, fast LP/SDP solvers, advances in interior point methods and fast streaming/sketching techniques. One highlighted theme was how combining the continuous and discrete points of view can often achieve near-optimal running time bounds.

Lee gave a talk at this workshop, Drusvyatskiy was one of the attendants, and several ADSI affiliated students also attended.