Nonsymmetric Search Directions for Semidefinite Programming

Two nonsymmetric search directions for semideenite programming, the XZ and ZX search directions, are proposed. They are derived from a nonsymmetric formulation of the semideenite programming problem. The XZ direction corresponds to the direct linearization of the central path equation XZ = I; while the ZX direction corresponds to ZX = I. The XZ and ZX directions are well deened if both X and Z are positive deenite matrices, where X may be nonsymmetric. We present an algorithm using the XZ and ZX directions alternately following the Mehrotra predictor-corrector framework. Numerical results show that the XZ/ZX algorithm is, in most cases, faster than the XZ+ZX method of Alizadeh, Overton, and Haeberly (AHO) while achieving similar accuracy.