Alternating proximal algorithms for constrained variational inequalities. Application to domain decomposition for PDE’s

Let $\cX,\cY,\cZ$ be real Hilbert spaces, let $f : \cX \rightarrow \R\cup\{+\infty\}$, $g : \cY \rightarrow \R\cup\{+\infty\}$ be closed convex functions and let $A : \cX \rightarrow \cZ$, $B : \cY \rightarrow \cZ$ be linear continuous operators. Let us consider the constrained minimization problem $$ \min\{f(x)+g(y):\quad Ax=By\}.\leqno (\cP)$$ Given a sequence $(\gamma_n)$ which tends toward $0$ as $n\to+\infty$, we study the following alternating proximal algorithm $$ \left\{ \begin{aligned} x_{n+1}&=\argmin\Big\{\gamma_{n+1}\,f(\zeta) + \frac{1}{2}\|A\zeta – By_n\|_\cZ^2 +\frac{\alpha}{2}\|\zeta – x_n\|_\cX^2; \,\,\, \zeta\in\cX\Big\}\\ y_{n+1}&=\argmin\Big\{\gamma_{n+1}\,g(\eta) + \frac{1}{2}\|Ax_{n+1} – B\eta\|_\cZ^2 +\frac{\nu}{2}\|\eta – y_n\|_\cY^2; \,\,\, \eta\in\cY\Big\}, \end{aligned} \right.\leqno (\cA) $$ where $\alpha$ and $\nu$ are positive parameters. It is shown that if the sequence $\left({\gamma_n}\right)$ tends {\em moderately slowly} toward $0$, then the iterates of $(\cA)$ weakly converge toward a solution of $(\cP)$. The study is extended to the setting of maximal monotone operators, for which a general ergodic convergence result is obtained. Applications are given in the area of domain decomposition for PDE’s.

Citation

Unpublished, Université Montpellier II, Universidad Técnica Federico Santa María.

Article

Download

View PDF