Struct GeometricProgramSolver 

pub struct GeometricProgramSolver {
    pub max_iter: usize,
    pub step_size: f64,
    pub tol: f64,
}

Geometric program solver via convex transformation.

A GP in standard form: minimize p_0(x) subject to p_i(x) ≤ 1, i=1..m, where each p_i is a posynomial. Under the change of variables x = exp(y), the GP becomes convex (log-sum-exp minimization).

This solver applies gradient descent on the log-domain objective.

Fields

max_iter: usize

Maximum iterations for the inner gradient descent.

step_size: f64

Step size for gradient descent in log-domain.

tol: f64

Convergence tolerance.

Implementations

impl GeometricProgramSolver

pub fn new(max_iter: usize, step_size: f64, tol: f64) -> Self

Create a new GP solver.

pub fn eval_log_monomial(log_c: f64, exponents: &[f64], y: &[f64]) -> f64

Evaluate a monomial c · ∏ x_i^{a_i} at log-domain point y (x = exp(y)). Returns log(c) + a^T y.

pub fn log_sum_exp_posynomial(monomials: &[(f64, Vec<f64>)], y: &[f64]) -> f64

Evaluate log of a posynomial: log(∑_k exp(log_c_k + a_k^T y)).

pub fn solve( &self, objective: &[(f64, Vec<f64>)], constraints: &[Vec<(f64, Vec<f64>)>], y0: &[f64], ) -> (Vec<f64>, f64)

Solve the GP: minimize objective posynomial subject to constraint posynomials ≤ 1.

objective: list of (log_coefficient, exponent_vector) pairs for objective posynomial. constraints: list of posynomials, each a list of (log_c, exponents) pairs.

Returns the optimal y = log(x) and the optimal objective value.

Trait Implementations

impl Clone for GeometricProgramSolver

fn clone(&self) -> GeometricProgramSolver

Returns a duplicate of the value.

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

Performs copy-assignment from source.

impl Debug for GeometricProgramSolver

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

Formats the value using the given formatter.