No label defined (Q4517249)

Let \(K\) be a nonempty closed convex set in \(\mathbb R^N\), let \(f\: K\to\mathbb R\) be a convex but not necessarily differentiable function, and let \(G\: K\to\mathbb R^N\) be a mapping. The mixed variational inequality problem is the problem of finding \(x^*\in K\) such that \[ \langle G(x^*),x-x^*\rangle + f(x) - f(x^*) \geq 0\quad\text{for all~}x\in K. \] Such problems have several applications in mathematical physics, economics, and operation research. Most articles were devoted to the case in which \(G\) is strictly monotone. The existing methods for solving the problem were designed for the case in which \(G\) is an \(M\)-matrix and \(K = \mathbb R^N\). In the article under consideration the above-mentioned problem is studied under different assumptions. It is supposed that \(G\) is a \(P\)-mapping and \(K\) is defined by box-type constraints. The aim of the article is to represent an approach for solving the mixed variational inequalities under the above more general conditions and then to prove global convergence of the corresponding method. The author considers a typical gap function and shows that it allows us to convert the initial mixed variational inequality into a constrained non-differentiable optimization problem as in the case of a strongly monotone \(G\). This function enables us to transform the typical variational inequality including a continuously differentiable and strongly monotone mapping into an unconstrained optimization problem. Thus, it is shown that the original problem can be reduced to a constrained optimization problem for a conventional gap function which may be a nondifferentiable function. At the same time, the \(D\)-gap function allows us to reduce the original problem to the problem of finding stationary points of a continuously differentiable function which can be solved by standard unconstrained differentiable optimization methods.