Bordered linear solvers (BLS)

The bordered linear solvers must be subtypes of AbstractBorderedLinearSolver <: AbstractLinearSolver.

The methods provided here solve bordered linear equations. More precisely, one is interested in the solution $u$ to $J\cdot u = v$ where

\[\tag E J=\left(\begin{array}{ll} {A} & {b} \\ {c^T} & {d} \end{array}\right) \text { and } v=\left(\begin{array}{l} {v_1} \\ {v_2} \end{array}\right)\]

Such linear solver bdlsolve will be called like sol, success, itnumber = bdlsolve(A, b, c, d, v1, v2) throughout the package.

Complex numbers

In the case where $c\in\mathbb C^N$, please note that the adjoint operator $c^T$ involves a conjugate.

Full matrix MatrixBLS

This easiest way to solve $(E)$ is by forming the matrix $J$. In case it is sparse, it should be relatively efficient. You can create such bordered linear solver using bls = MatrixBLS(ls) where ls::AbstractLinearSolver is a linear solver (which defaults to \) used to solve the linear problem associated to $J$. This is the default method used in the package.

Bordering method BorderingBLS

The general solution $u=\left(\begin{array}{l} {u_1} \\ {u_2} \end{array}\right)$ to $(E)$ when $A$ is non singular is

\[\begin{array}{l} u_2 = \frac{1}{d - c\cdot x_2}(v_2 - c\cdot x_1) \\ u_1=x_1-u_2x_2 \end{array}\]

where $x_1=A^{-1}v_1, x_2=A^{-1}b$.

It is very efficient for large scale problems because it is entirely Matrix-Free and one can use preconditioners. You can create such bordered linear solver using bls = BorderingBLS(ls) where ls::AbstractLinearSolver is a linear solver which defaults to \. The intermediate solutions $x_1=A^{-1}v_1, x_2=A^{-1}b$ are formed using ls.

  1. In the case where ls = DefaultLS(), the factorisation of A is cached so the second linear solve is very fast

There are more options to BorderingBLS. First, the residual can be checked using the option check_precision = true. If the residual is above a prescribed tolerance, an iterative method is used based on several bordering transformations. This is the BEC+k algorithm in [Govaerts].

Full Matrix-Free MatrixFreeBLS

In cases where $A$ is singular but $J$ is not, the bordering method may fail. It can thus be advantageous to form the Matrix-Free version of $J$ and call a generic linear solver to find the solution to $(E)$. You can create such bordered linear solver using bls = MatrixFreeBLS(ls) where ls::AbstractLinearSolver is a (Matrix Free) linear solver which is used to invert J.

For now, this linear solver only works with AbstractArray

References

  • Govaerts

    Govaerts, W. “Stable Solvers and Block Elimination for Bordered Systems.” SIAM Journal on Matrix Analysis and Applications 12, no. 3 (July 1, 1991): 469–83. https://doi.org/10.1137/0612034.