Equivalent differentiable optimization problems and descent methods for asymmetric variational inequality problems (Q1184339): Difference between revisions

The variational inequality problem is (1): Find \(x^*\in S\subset\mathbb{R}^ n\) such that \(\langle F(x^*),x-x^*\rangle\geq 0\), \(\forall x\in\mathbb{R}^ n\) where \(S\neq\emptyset\) is closed and convex and \(F: \mathbb{R}^ n\to\mathbb{R}^ n\). Now with \(G\) being any \(n\times n\) positive definite matrix consider the program (2): \(\min\{\phi(y): y\in S\}\) where \(\phi(y)=\langle F(x),(y-x)\rangle+{1\over 2}\langle(y-x),G(y-x)\rangle\), and let \(-f(x): \mathbb{R}^ n\to\mathbb{R}\) be the optimal objective value of \(\phi(y)\) in (2). The main result is now: (i) \(f(x)\geq 0\), \(\forall x\in S\), and (ii) \(x^*\) solves (1) if and only if it solves the program (3): \(\min\{f(x): x\in S\}\) and that happens if and only if \(f(x^*)=0\), \(x^*\in S\). Moreover \(f\) is continuously differentiable (continuous) if \(F\) is continuously differentiable (continuous). In the first case descent methods are presented to solve the program (3) by an iteration process. A list of sixteen references closes the paper.