[tag] anderson

Configuration of the Anderson acceleration for the Picard iteration.

Anderson acceleration mixes the last \( m \) (damped) Picard steps instead of taking the most recent one only, which typically reduces the number of iterations for slowly converging fixed-point problems.

Example:**

<anderson>
    <depth>3</depth>
</anderson>

Omitting the anderson subtree disables the acceleration and gives a plain (possibly damped) Picard iteration.

Child parameters, attributes and cases

• [tag] depth

Additional info

From NumLib/ODESolver/CreateNonlinearSolver.cpp line 38

• This is an optional parameter.
• Expanded tag path: nonlinear_solvers.nonlinear_solver.Picard.anderson
• Go to source code: → ogs/ogs/6.5.9

Used in the following test data files

• [→ ogs/ogs/master | → doc] Parabolic/ComponentTransport/HTCWithFracture/2D_single_fracture_HTC.prj
• [→ ogs/ogs/master | → doc] Parabolic/T/1D_freezing_column_Stefan/Stefan_problem.prj