Interior proximal algorithm for quasiconvex programming problems and variational inequalities with linear constraints (Q1762409)

For solving general linearly constrained convex programming problems, many works proposed algorithms based on a function that is constructed by adding a quadratic regularization term to the barrier term that enforces the generated sequence to remain in the interior of the feasible region. In this paper, the authors extend this method first to quasiconvex minimization problems and then to quasimonotone variational inequalities. In that purpose, they replace the quadratic regularization term by a distance-like function and they prove the global convergence of the resulting methods. For quasiconvex minimization problems, they assume that the regularization parameters go to zero, which is not necessary when the objective function is pseudoconvex. For variational inequality problems, the convergence is obtained under the assumption that the operator is quasimonotone, which is a weak condition. As mentioned by the authors, the proposed schemes for solving the two problems are mainly theoretical.