Projection and contraction methods for semidefinite programming

摘要

Recently, He (1992, 1994a,b, 1996) presented some iterative projection and contraction (PC) methods for linear variational inequalities. Its simplicity, robustness and ability to handle large problems has attracted a lot of attention. This paper extends the PC methods from linear programming to semidefinite programming (SDP). The SDP problem is transformed into an equivalent projection equation, which is solved by PC methods. Some promising numerical results obtained from a preliminary implementation are included.