Nonsmooth Algorithms to Solve Semidefinite Programs

Semi-definite programming (SDP) is of growing importance in various filed: system control, mechanics, combinatorial optimization, ... Usually, it is solved by interior point methods, which are elegant, efficicent and well-suited. However, they have limitations, particularly in large-scale or ill-conditioned cases. On the other hand, SDP is an instance of non-smooth optimization (NSO), which enjoys some particular structure. We briefly review the works that have been devoted to solving SDP with NSO tools, and we outline the possibilities for future work.