Struct newton_rootfinder::solver::SolverParameters

pub struct SolverParameters { /* fields omitted */ }

A minimal struct holding the resolution parameters

Parameters

Damping

Activate the damping to improve convergence

Plain resolution according to Newton is made through the formula X = X - J^-1*F(X)

However, if the proposed update is not performing (deterioriating the solution) it is likely it is due to a much to step-size too important. Reducing the step-size might be the solution

The damping formula is then : X = X - damping_factorJ^-1F(X) with 0 < damping_factor <= 1

As long as the error is reduced damping_factor = 1. If it is not the case, a factor is applied (the value might change according to the versions).

If the used method is a quasi-newton method and the jacobian has not been updated at the current iteration, damping won’t be performed but the jacobian will be recomputed at the next iteration.

In the case of the jacobian has been recomputed at the current iteration, damping will be performed

Tolerance

The tolerance values used by the solver to check for convergence.

Each residuals must be below this threshold

Max iteration

The maximum number of iterations the solver is allowed to make

This is required to avoid to have an infinte loop

Resolution Method

See the enum ResolutionMethod for the options available

Problem size

The dimension of the problem for the resolution

Implementations

impl SolverParameters

pub fn new(
    problem_size: usize,
    tolerance: f64,
    max_iter: usize,
    resolution_method: ResolutionMethod,
    damping: bool
) -> Self

pub fn get_problem_size(&self) -> usize

pub fn get_tolerance(&self) -> f64

pub fn get_max_iter(&self) -> usize

pub fn get_resolution_method(&self) -> ResolutionMethod

pub fn get_damping(&self) -> bool

Trait Implementations

impl Display for SolverParameters

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.