yaeos__math_nonlinearsolvers Module

Nonlinear equation solvers

Description

Newton-type solvers for systems of nonlinear equations, including the damped multivariate Newton method newton. newton is the damped Newton method; homotopy gradually deforms an identity problem into the target system.

Examples

call newton(residual_jacobian, x, tol, max_its, iterations)

Uses

  • module~~yaeos__math_nonlinearsolvers~~UsesGraph module~yaeos__math_nonlinearsolvers yaeos__math_nonlinearsolvers module~yaeos__constants yaeos__constants module~yaeos__math_nonlinearsolvers->module~yaeos__constants module~yaeos__math_linalg yaeos__math_linalg module~yaeos__math_nonlinearsolvers->module~yaeos__math_linalg iso_fortran_env iso_fortran_env module~yaeos__constants->iso_fortran_env module~yaeos__math_linalg->module~yaeos__constants

Used by

  • module~~yaeos__math_nonlinearsolvers~~UsedByGraph module~yaeos__math_nonlinearsolvers yaeos__math_nonlinearsolvers proc~solve_tp solve_TP proc~solve_tp->module~yaeos__math_nonlinearsolvers

Interfaces

public interface to_solve

  • public subroutine to_solve(X, F, J)

    Callback interface for a square nonlinear system and dense Jacobian.

    Arguments

    Type IntentOptional Attributes Name
    real(kind=pr), intent(in) :: X(:)

    Current unknown vector.

    real(kind=pr), intent(out) :: F(:)

    Residual vector .

    real(kind=pr), intent(out) :: J(:,:)

    Jacobian .


Subroutines

public subroutine homotopy(sub, x, tol, max_its, its)

Attempt a homotopy solve from an easy identity system to the target.

Read more…

Arguments

Type IntentOptional Attributes Name
procedure(to_solve) :: sub

Target residual/Jacobian callback.

real(kind=pr), intent(inout) :: x(:)

Initial guess, updated homotopy solution.

real(kind=pr), intent(in) :: tol

Absolute residual tolerance at each stage.

integer, intent(in) :: max_its

Present for interface compatibility; current implementation uses a fixed stage limit.

integer, intent(out) :: its

Iterations reported by the last Newton stage.

public subroutine newton(sub, x, tol, max_its, its)

Multidimensional Newton method with simple step damping.

Read more…

Arguments

Type IntentOptional Attributes Name
procedure(to_solve) :: sub

Callback returning residuals and the Jacobian.

real(kind=pr), intent(inout) :: x(:)

Initial guess, overwritten by the current solution estimate.

real(kind=pr), intent(in) :: tol

Absolute infinity-norm tolerance on the residual.

integer, intent(in) :: max_its

Maximum number of Newton iterations.

integer, intent(out) :: its

Iteration count at exit.