A. M.-C. So
Many intractable problems in engineering can be formulated as a semidefinite program (SDP) with a rank constraint. Currently, a standard approach to tackle these problems is semidefinite relaxation. The idea is to drop the rank constraint to get an efficiently solvable SDP. However, standard SDP solvers typically yield high-rank solutions. In this project, we investigate the use of nonconvex regularization terms to promote low-rank solutions. The focus will be on both the computational complexity of such approaches and their practical implementations.