Trait optimization_engine::core::AlgorithmEngine

pub trait AlgorithmEngine {
    fn step(&mut self, _: &mut [f64]) -> bool;
    fn init(&mut self, _: &mut [f64]);
}

Engine supporting an algorithm

An engine is responsible for the allocation of memory for an algorithm, especially memory that is reuasble is multiple instances of the same algorithm (as in model predictive control).

It defines what the algorithm does at every step (see step) and whether the specified termination criterion is satisfied

Required methods

fn step(&mut self, _: &mut [f64]) -> bool

Take a step of the algorithm and return true only if the iterations should continue

fn init(&mut self, _: &mut [f64])

Loading content...

Implementors

impl<'a, GradientType, ConstraintType, CostType> AlgorithmEngine for FBSEngine<'a, GradientType, ConstraintType, CostType> where
    GradientType: Fn(&[f64], &mut [f64]) -> i32 + 'a,
    CostType: Fn(&[f64], &mut f64) -> i32 + 'a,
    ConstraintType: Constraint + 'a, 

fn step(&mut self, u_current: &mut [f64]) -> bool

Take a forward-backward step and check whether the algorithm should terminate

Arguments

• u_current the current mutable

Returns

• A boolean flag which istrue if and only if the algorith should not terminate

Panics

The method may panick if the computation of the gradient of the cost function or the cost function panics.

fn init(&mut self, _u_current: &mut [f64])

impl<'a, GradientType, ConstraintType, CostType> AlgorithmEngine for PANOCEngine<'a, GradientType, ConstraintType, CostType> where
    GradientType: Fn(&[f64], &mut [f64]) -> i32,
    CostType: Fn(&[f64], &mut f64) -> i32,
    ConstraintType: Constraint, 

Implementation of the step and init methods of [trait.AlgorithmEngine.html]

fn step(&mut self, u_current: &mut [f64]) -> bool

PANOC step

Performs a step of PANOC, including the line search

Arguments

• u_current on entry is the current iterate; on exit, it is updated with the next iterate of PANOC

fn init(&mut self, u_current: &mut [f64])

Initialization of PANOC

Computes a number of essential quantities before the start of PANOC iterations

There include the computation of the cost and gradient of the cost at the initial point, the computation of an initial estimation of the Lipschitz constant of the gradient of the cost at the initial point, initial estimates for gamma and sigma, a gradient step and a half step (projected gradient step)

Loading content...