Equivalent Unconstrained Minimization and Global Error Bounds for Variational Inequality Problems

From MaRDI portal





The paper deals with constructions of merit functions for a variational inequality \(\langle F(x),y-x\rangle \geq 0\) for all \(y\in S\), where \(F\) is a continuous mapping from \(R^n\) into itself and \(S\) is a closed convex subset of \(R^n.\) It is concerned with the following functions derived from the Moreau-Yosida regularization: \[ \varPhi_f(x;\alpha,\lambda)=\inf_{z\in S}\{f(z;\alpha)+\lambda|x-z|^2\},\;\varPhi_h(x;\beta,\lambda)=\inf_{z\in S}\{h(z;\beta)+\lambda|x-z|^2\}, \] where \(\lambda>0\) and \[ f(x;\alpha)=\sup_{y\in S} \{\Psi(x,y)-\alpha|x-y|^2\},\;h(x;\beta)=\sup_{y\in S} \{-\Psi(y,x)+\beta|x-y|^2\}, \] \(\alpha\geq 0,\;\beta\geq 0,\;\Psi(x,y)=\langle F(x),x-y\rangle.\) The equivalence between the considered variational inequality and the unconstrained minimization of \(\varPhi_f(.;\alpha,\lambda)\) and \(\varPhi_h(.;\beta,\lambda)\) is established. Error bounds for variational inequalities using the merit functions are derived.




Cited in
(61)








This page was built for publication: Equivalent Unconstrained Minimization and Global Error Bounds for Variational Inequality Problems

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q3128460)