Struct ProximalGradientOptimizer 

pub struct ProximalGradientOptimizer {
    pub max_iter: usize,
    pub tol: Float,
    pub step_size: Option<Float>,
    pub use_line_search: bool,
    pub accelerated: bool,
    pub beta: Float,
    pub sigma: Float,
}

Proximal Gradient Method optimizer for composite objectives

Fields

max_iter: usize

Maximum number of iterations

tol: Float

Convergence tolerance

step_size: Option<Float>

Initial step size

use_line_search: bool

Whether to use line search for step size

accelerated: bool

Whether to use acceleration (FISTA)

beta: Float

Backtracking line search parameters

sigma: Float

Implementations

impl ProximalGradientOptimizer

pub fn accelerated() -> Self

Create accelerated version (FISTA)

pub fn minimize<F, GradF, ProxG>( &self, f: F, grad_f: GradF, prox_g: ProxG, x0: Array1<Float>, ) -> Result<Array1<Float>, String>

where F: Fn(&Array1<Float>) -> Float, GradF: Fn(&Array1<Float>) -> Array1<Float>, ProxG: Fn(&Array1<Float>, Float) -> Array1<Float>,

Minimize composite objective f(x) + g(x) using proximal gradient method where f is smooth and g is convex (possibly non-smooth)

Trait Implementations

impl Default for ProximalGradientOptimizer

fn default() -> Self

Returns the “default value” for a type.