Variational optimization of iterative decomposition methods

Iterative methods are studied for solving a system of linear algebraic equations \(Ax=b\) which arises in mathematical physics. The main problem is to determine a sequence of the parameters \(\omega\), \(\tau\) in the ``generating scheme \[ B_{m+1}(\omega^{\alpha}_{m+1}) \frac{x^{m+1}-x^{m}}{\tau_{m+1}}+Ax_{m}=b, \] \[ B_{m+1}=(I+\omega^1_{m+1}A_1)(I+\omega^2_{m+1}A_2), \] where the matrix \(A\) is such that \[ A=A_1+A_2, A_{\alpha}=A_{\alpha}^T>0, A_1A_2=A_2A_1, 0<\delta_{\alpha}I<A<\Delta_{\alpha}I. \] The Douglas-Rachford and Pieceman-Rachford methods follow from the above general scheme. A variational approach is proposed to determine an explicit form of optimal iterative parameters \(\omega^{\alpha}_{m+1}\), \(\tau_{m+1}\) via minimization of some functional. A proof of convergence and an estimate of the rate of convergence are given. The author mentions that some of the obtained results remain valid under more general conditions, i.e., when the matrices \(A_{\alpha}\) are noncommutative. He intends to publish them in the future.

Recommendations

Cites work

Cited in

(3)

This page was built for publication: Variational optimization of iterative decomposition methods

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q1360840)