A new primal-dual path-following interior-point algorithm for semidefinite optimization (Q1014697)

A new class of search directions, proposed in the context of interior-point methods for linear programming (LP), is extended to semidefinite programming (SDP), a generalization of linear programming to the convex cone of positive semidefinite matrices. Closely following the LP convergence analysis developed for a primal-dual path-following algorithm, the authors prove similar iteration bounds for SDP, in the case of small updates. Unfortunately, this theoretical iteration count is not illustrated on numerical examples and no implementation of this algorithm seems to be available.