Uzawa iteration
In
Basic idea
We consider a saddle point problem of the form
where is a symmetric
where denotes the Schur complement. Since is symmetric positive-definite, we can apply standard iterative methods like the gradient descent method or the conjugate gradient method to
in order to compute . The vector can be reconstructed by solving
It is possible to update alongside during the iteration for the Schur complement system and thus obtain an efficient algorithm.
Implementation
We start the conjugate gradient iteration by computing the residual
of the Schur complement system, where
denotes the upper half of the solution vector matching the initial guess for its lower half. We complete the initialization by choosing the first search direction
In each step, we compute
and keep the intermediate result
for later. The scaling factor is given by
and leads to the updates
Using the intermediate result saved earlier, we can also update the upper part of the solution vector
Now we only have to construct the new search direction by the Gram–Schmidt process, i.e.,
The iteration terminates if the residual has become sufficiently small or if the norm of is significantly smaller than indicating that the Krylov subspace has been almost exhausted.
Modifications and extensions
If solving the linear system exactly is not feasible, inexact solvers can be applied.[2][3][4]
If the Schur complement system is ill-conditioned, preconditioners can be employed to improve the speed of convergence of the underlying gradient method.[2][5]
Inequality constraints can be incorporated, e.g., in order to handle obstacle problems.[5]
References
- ^ Uzawa, H. (1958). "Iterative methods for concave programming". In Arrow, K. J.; Hurwicz, L.; Uzawa, H. (eds.). Studies in linear and nonlinear programming. Stanford University Press.
- ^ doi:10.1137/0731085.
- .
- .
- ^ ISBN 978-3-540-34468-1.
Further reading
- Chen, Zhangxin (2006). "Linear System Solution Techniques". Finite Element Methods and Their Applications. Berlin: Springer. pp. 145–154. ISBN 978-3-540-28078-1.