This class describes a More-Thuente line search as presented in the paper "Line Search Algorithms with Guaranteed Sufficient Decrease" by Jorge J. More and David J. Thuente. More...

#include <more_thuente_line_search.hpp>

Inheritance diagram for autoware::common::optimization::MoreThuenteLineSearch:

[legend]

Collaboration diagram for autoware::common::optimization::MoreThuenteLineSearch:

[legend]

Public Types
enum	OptimizationDirection { OptimizationDirection::kMinimization, OptimizationDirection::kMaximization }
	An enum that defines the direction of optimization. More...

Public Types inherited from autoware::common::optimization::LineSearch< MoreThuenteLineSearch >
using	StepT = float_t

Public Member Functions
	MoreThuenteLineSearch (const StepT max_step, const StepT min_step, const OptimizationDirection optimization_direction=OptimizationDirection::kMinimization, const StepT mu=1.e-4F, const StepT eta=0.1F, const std::int32_t max_iterations=10)
	Constructs a new instance. More...

template<typename DomainValueT , typename OptimizationProblemT >
DomainValueT	compute_next_step_ (const DomainValueT &x0, const DomainValueT &initial_step, OptimizationProblemT &optimization_problem)
	Calculates the next step. More...

Public Member Functions inherited from autoware::common::optimization::LineSearch< MoreThuenteLineSearch >
	LineSearch (const StepT step_max=std::numeric_limits< StepT >::min())

DomainValueT	compute_next_step (const DomainValueT &x0, const DomainValueT &initial_step, OptimizationProblemT &optimization_problem)

StepT	get_step_max () const noexcept

void	set_step_max (const StepT step_max) noexcept

Additional Inherited Members
Protected Member Functions inherited from autoware::common::helper_functions::crtp< MoreThuenteLineSearch >
const MoreThuenteLineSearch &	impl () const

MoreThuenteLineSearch &	impl ()

Detailed Description

This class describes a More-Thuente line search as presented in the paper "Line Search Algorithms with Guaranteed Sufficient Decrease" by Jorge J. More and David J. Thuente.

One notable change here is that this class will automatically detect if we are trying to solve a minimization or the maximization problem by evaluating \(\phi^\prime(0)\). If \(\phi^\prime(0) < 0\) then we follow the suggestions in the paper exactle. Otherwise, we assume we are dealing with a dual maximization problem and will flip the objective function \(\phi\), thus making it again a minimization problem.

Member Enumeration Documentation

◆ OptimizationDirection

enum autoware::common::optimization::MoreThuenteLineSearch::OptimizationDirection

strong

An enum that defines the direction of optimization.

Enumerator
kMinimization
kMaximization

Constructor & Destructor Documentation

◆ MoreThuenteLineSearch()

autoware::common::optimization::MoreThuenteLineSearch::MoreThuenteLineSearch	(	const StepT	max_step,
		const StepT	min_step,
		const OptimizationDirection	optimization_direction = `OptimizationDirection::kMinimization`,
		const StepT	mu = `1.e-4F`,
		const StepT	eta = `0.1F`,
		const std::int32_t	max_iterations = `10`
	)

inlineexplicit

Constructs a new instance.

Parameters

[in]	max_step	The maximum step that the process is allowed to make
[in]	min_step	The minimum step that the process is allowed to make
[in]	optimization_direction	The optimization direction
[in]	mu	Constant \(\mu\) (eq. 1.1), that forces a sufficient decrease of the function.
[in]	eta	Constant \(\eta\) (eq. 1.2), that forces the curvature condition.
[in]	max_iterations	Maximum allowed iterations.

Member Function Documentation

◆ compute_next_step_()

template<typename DomainValueT , typename OptimizationProblemT >

DomainValueT autoware::common::optimization::MoreThuenteLineSearch::compute_next_step_	(	const DomainValueT &	x0,
		const DomainValueT &	initial_step,
		OptimizationProblemT &	optimization_problem
	)

Calculates the next step.

Parameters

[in]	x0	Starting argument.
[in]	initial_step	Initial step to initiate the search.
	optimization_problem	The optimization problem for generating values of function denoted as f in the paper.

Template Parameters

DomainValueT	Type of values of the domain of the function f.
OptimizationProblemT	The type of the optimization problem that provides values of function f evaluated on a domain point.

Returns: The new step to make in order to optimize the function.

Note: If the length of initial_step is smaller than the m_step_min then an unmodified initial_step will be returned.

The documentation for this class was generated from the following file:

more_thuente_line_search.hpp

Public Types

Public Member Functions