MinimumResidualSolver<T> Class

public sealed class MinimumResidualSolver<T> : IterativeSparseSolver<T>

Public NotInheritable Class MinimumResidualSolver(Of T)
	Inherits IterativeSparseSolver(Of T)

generic<typename T>
public ref class MinimumResidualSolver sealed : public IterativeSparseSolver<T>

[<SealedAttribute>]
type MinimumResidualSolver<'T> = 
    class
        inherit IterativeSparseSolver<'T>
    end

Inheritance

Object  →  ManagedIterativeAlgorithm<DenseVector<T>, T, SolutionReport<DenseVector<T>, T>>  →  IterativeSparseSolver<T>  →  MinimumResidualSolver<T>

Type Parameters

T

The real or complex floating-point scalar type.

The operator may be indefinite. Explicit matrices must be Hermitian; callers must ensure that matrix-free operators are Hermitian. Identity preconditioning is used by default. A supplied Preconditioner must remain a fixed linear, Hermitian positive-definite operator throughout each solve. The inherited LeftPreconditioner is not applied.

With a supplied preconditioner, residual minimization uses its induced norm. Stopping criteria still use the residual of the original system.

Convergence uses the true residual of the original system. Compatible singular systems can converge, but no minimum-length solution is promised. Lanczos breakdown with an unsatisfied residual, or an observed invalid preconditioner metric, reports RoundOffError. Arbitrary operator and preconditioner symmetry or definiteness is not certified.

Reference