In this paper, using the framework of self-regularity, we propose a hybrid adaptive algorithm for the linear optimization problem. If the current iterates are far from a central path, the algorithm employs a self-regular search direction, otherwise the classical Newton search direction is employed. This feature of the algorithm allows us to prove a worst case iteration bound. Our result matches the best iteration bound obtained by the pure self-regular approach and improves on the worst case iteration bound of the classical algorithm.

We introduce a new barrier function to solve a class ofSemidefinite Optimization Problems (SOP) with bounded variables.That class is motivated by some (SOP) as the minimization of thesum of the first few eigenvalues of symmetric matrices and graphpartitioning problems. We study the primal-dual central pathdefined by the new barrier and we show that this path is analytic,bounded and that all cluster points are optimal solutions of theprimal-dual pair of problems. Then, using some ideas fromsemi-analytic geometry we prove its full convergence. Finally, weintroduce a new proximal point algorithm for that class ofproblems and prove its convergence.

Les méthodes de points intérieurs en programmation linéaireconnaissent un grand succès depuis l'introductionde l'algorithme de Karmarkar. La convergence de l'algorithme repose sur unefonction potentielle qui, sous saforme multiplicative, fait apparaître un exposant p. Cet exposantest, de façongénérale, choisi supérieur au nombre de variables n du problème.Nous montrons dans cetarticle que l'on peut utiliser des valeurs dep plus petites que n. Ceci permet d'améliorer le conditionnement dela méthode au voisinage de la solution optimale.

In this note, we present a slight modification of an algorithm for thestrict feasibility problem. This modification reducesthe number of iterations.