Abstract (100 words or less): |  | Interior point methods (IPMs) can solve large scale LPs and SOCPs, however, they are fairly limited in the size of SDPs they can handle. Extending the work of Krishnan and Mitchell (2001), we recently proposed a conic interior point decomposition approach to improve the scalability of IPMs for solving large scale SDPs in practice. In this talk, we motivate and develop the decomposition approach for SDP, and also highlight some of the issues involved in an efficient implementation of the algorithm. Some of our computational experiences with this algorithm will also be presented. |