Struct ProximalBundle 

pub struct ProximalBundle {
    pub options: ProximalBundleOptions,
}

Proximal bundle method for nonsmooth convex optimization.

Solves min f(x) where f is convex but possibly nonsmooth. Uses a bundle of subgradients to build a cutting-plane model:

f̂(x; y) = max_{i in B} { f(x_i) + g_i^T (x - x_i) }

The proximal subproblem:

x^{k+1} = argmin_x { f̂(x; y^k) + (μ/2) ||x - y^k||² }

generates either a serious step (sufficient decrease) or null step.

Fields

options: ProximalBundleOptions

Algorithm options

Implementations

impl ProximalBundle

pub fn new(options: ProximalBundleOptions) -> Self

Create a proximal bundle solver

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

where FS: Fn(&[f64]) -> (f64, Vec<f64>),

Solve min f(x) using the proximal bundle method.

Arguments

• func - Objective function (can be nonsmooth)
• subgrad - Subgradient of f at x: returns (f(x), g ∈ ∂f(x))
• x0 - Initial point

Trait Implementations

impl Default for ProximalBundle

fn default() -> Self

Returns the “default value” for a type.