Central paths in semidefinite programming, generalized proximal-point method and Cauchy trajectories in Riemannian manifolds (Q1014028)

In the context of semidefinite programming, the authors prove the equivalence between three concepts, namely, the central path, the generalized proximal-point sequence and the Cauchy trajectory in Riemannian manifold. These results are natural extensions of results obtained by Iusem et al. First the convergence properties of the central path are studied when it is associated with a real-valued function defined on the set of all positive definite symmetric matrices. In particular when this function can be continuously extended to the boundary of its domain, the central path is well defined and bounded, and is converging to the analytical center of the solution set of the problem. Using a generalized distance associated with this function in the proximal-point method, the authors prove that the correspondingly generated sequence is contained in the central path and is converging to the analytical center. Finally the relationship between the Cauchy trajectory in Riemannian manifold and the central path is established.