A class of \(D\)-gap functions for mixed variational inequalities (Q1589138)

Let \(U\) be a nonempty closed convex subset of \(R^n\), \(G: R^n \to R^n\) a continuously differentiable mapping, \(F: R^n \to R\) a convex and continuous function. The mixed variational inequality problem is the problem of finding \(u^* \in U\) such that \[ \langle G(u^*), u-u^* \rangle+ f(u)-f(u^*) \geq 0\quad \forall u \in U. \] It is shown that the problem is equivalent to the unconstrained differentiable optimization problem \[ \psi_{\alpha\beta}(u) \to \min, \] where \(\psi_{\alpha\beta}(u)=\varphi_{\alpha}(u)-\varphi_{\beta}(u)\), \(0<\alpha<\beta\) and \[ \varphi_{\alpha}(u)=\max_{u \in U} \Phi_{\alpha}(u,v), \] \[ \Phi_{\alpha}(u,v)=\langle G(u), u-v\rangle - 0.5 \alpha \|u-v \|^2+ f(u)-f(v). \] An iterative method for solving this problem is presented.