Struct MultigridOptimizer 

pub struct MultigridOptimizer<F>
where
    F: Fn(&[f64]) -> f64,
{
    pub func: F,
    pub n_fine: usize,
    pub n_levels: usize,
    pub options: MultigridOptions,
}

Multigrid-inspired optimizer using a hierarchy of approximation levels.

The “grid levels” here represent different approximation granularities. Restriction maps a fine iterate to a coarser space; prolongation (interpolation) maps the coarse correction back to the fine space.

On each smoothing step, gradient descent is applied. The multigrid philosophy avoids slow convergence of fine-level iterations by first making progress on coarser levels.

Fields

func: F

Objective function (same function evaluated at different scales)

n_fine: usize

Dimensionality at finest level

n_levels: usize

Number of multigrid levels

options: MultigridOptions

Algorithm options

Implementations

impl<F: Fn(&[f64]) -> f64> MultigridOptimizer<F>

pub fn new( func: F, n_fine: usize, n_levels: usize, options: MultigridOptions, ) -> Self

Create a new multigrid optimizer

pub fn minimize(&self, x0: &[f64]) -> OptimizeResult<MultilevelResult>

Run the multigrid optimizer