Symmetric primal-dual path-following algorithms for semidefinite programming

We propose a framework for developing and analyzing primal-dual interior point algorithms for semidefinite programming. This framework is an extension of the v-space approach that was developed by Kojima et al. (1991) for linear complementarity problems. The extension to semidefinite programming allows us to interpret Nesterov-Todd type directions (Nesterov and Todd 1995, 1997) as Newton search directions. Our approach does not involve any barrier function. Several primal-dual path-following algorithms for semidefinite programming are analyzed. The treatment of these algorithms for semidefinite programming in our setting bears great similarity to the linear programming case.