Struct argmin::backtracking::BacktrackingLineSearch

pub struct BacktrackingLineSearch<'a> { /* fields omitted */ }

Backtracking Line Search

Methods

impl<'a> BacktrackingLineSearch<'a>

fn new(
    cost_function: &'a Fn(&Array1<f64>) -> f64,
    gradient: &'a Fn(&Array1<f64>) -> Array1<f64>
) -> Self

Initialize Backtracking Line Search

Requires the cost function and gradient to be passed as parameter. The parameters max_iters, tau, and c are set to 100, 0.5 and 0.5, respectively.

fn alpha(&mut self, alpha: f64) -> &mut Self

Set the maximum distance from the starting point

fn max_iters(&mut self, max_iters: u64) -> &mut Self

Set the maximum number of iterations

fn c(&mut self, c: f64) -> Result<&mut Self, Error>

Set c to a desired value between 0 and 1

fn tau(&mut self, tau: f64) -> Result<&mut Self, Error>

Set tau to a desired value between 0 and 1

Trait Implementations

impl<'a> ArgminSolver<'a> for BacktrackingLineSearch<'a>

type Parameter = Array1<f64>

Parameter vector

type CostValue = f64

Cost value

type Hessian = Array1<f64>

Hessian

type StartingPoints = Array1<f64>

Initial parameter(s)

type ProblemDefinition = Array1<f64>

Type of Problem

fn init(
    &mut self,
    p: Self::ProblemDefinition,
    x: &Self::StartingPoints
) -> Result<(), Error>

Initializes the solver and sets the state to its initial state

fn next_iter(
    &mut self
) -> Result<ArgminResult<Self::Parameter, Self::CostValue>, Error>

Moves forward by a single iteration

fn terminate(&self) -> TerminationReason

Handles the stopping criteria

fn run(
    &mut self,
    operator: Self::ProblemDefinition,
    init_param: &Self::StartingPoints
) -> Result<ArgminResult<Self::Parameter, Self::CostValue>, Error>

Run initialization and iterations at once