Abstract

We generalize the projection method for solving strongly monotone multivalued variational inequalities when the cost operator is not necessarily Lipschitz. At each iteration at most one projection onto the constrained set is needed. When the convex constrained set is not polyhedral, we embed the proposed method in a polyhedral outer approximation procedure. This allows us to obtain the projections by solving strongly convex quadratic programs with linear constraints. We also discuss how to use the proposed method to implement inexact proximal point methods.