Struct ProximalGradientSolver 

pub struct ProximalGradientSolver {
    pub lipschitz: f64,
    pub max_iter: usize,
    pub tol: f64,
    pub accelerated: bool,
}

Proximal gradient method (ISTA/FISTA) for composite optimization.

Minimizes f(x) + g(x) where f is L-smooth (known Lipschitz constant) and g has a cheap prox operator. Supports both ISTA (beta=0) and FISTA (beta>0, Nesterov momentum).

Fields

lipschitz: f64

Lipschitz constant L of gradient of f.

max_iter: usize

Maximum number of iterations.

tol: f64

Convergence tolerance on successive iterate change.

accelerated: bool

Use FISTA acceleration (true) or plain ISTA (false).

Implementations

impl ProximalGradientSolver

pub fn new(lipschitz: f64, max_iter: usize, tol: f64, accelerated: bool) -> Self

Create a proximal gradient solver.

pub fn minimize<F, G>( &self, smooth: &F, regularizer: &G, x0: &[f64], ) -> (Vec<f64>, usize)

where F: ConvexFunction, G: ProxableFunction,

Minimize smooth + regularizer. Returns (x_star, iters).

pub fn estimate_lipschitz<F: ConvexFunction>( f: &F, x: &[f64], num_trials: usize, ) -> f64

Estimate the Lipschitz constant via power iteration on the Hessian approximation.

Returns an upper bound on L by computing max ||grad f(x + eps * v) - grad f(x)||/eps.

Trait Implementations

impl Clone for ProximalGradientSolver

fn clone(&self) -> ProximalGradientSolver

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for ProximalGradientSolver

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

Formats the value using the given formatter.