Scalable Rapidly Deployable Convex Optimization for Data Analytics

Over the period of the contract we have developed the full stack for wide use of convex optimization, in machine learning and many other areas. We have developed a new open-source domain specific language DSL for convex optimization, in Python, Julia, and recently, in R. Each of these has been published, and each are widely used and cited. We have developed a novel open-source solver, SCS, that is bundled with the DSLs, and solves any combination of linear programs, SOCPs, SDPs, exponential cone programs, and power cone programs. CVXPY supports basic methods for distributed optimization, on multiple heterogenous platforms. We have also done basic research in various application areas, using CVXPY, to demonstrate its usefulness. See attached report for publication information.