Public Types
typedef ParameterType::Scalar	Scalar

Public Member Functions
	GradientDescent (std::shared_ptr< LineSearch< ProblemType, ParameterType > > line_search, std::function< bool(Scalar, Scalar, size_t)> progress)

void	minimize (const ProblemType &problem, Eigen::Ref< ParameterType > x0)

Detailed Description

A very simple implementation of batch gradient descent.

Given a problem, a line search and a starting point it performs the following simple iteration.

Convergence is decided by another part of the program through injection of a callback.

TODO: Maybe bind the problem type through an interface

Constructor & Destructor Documentation

template<typename ProblemType , typename ParameterType >

ldaplusplus::optimization::GradientDescent< ProblemType, ParameterType >::GradientDescent	(	std::shared_ptr< LineSearch< ProblemType, ParameterType > >	line_search,
		std::function< bool(Scalar, Scalar, size_t)>	progress
	)

inline

Parameters

line_search	A line search method
progress	A callback that decides when the optimization has ended; it can also be used to get informed about the progress of the optimization

template<typename ProblemType , typename ParameterType >

void ldaplusplus::optimization::GradientDescent< ProblemType, ParameterType >::minimize	(	const ProblemType &	problem,
		Eigen::Ref< ParameterType >	x0
	)

inline

Minimize the function defined in the 'problem' argument.

The 'problem' argument should implement the functions value() and gradient() in order to be used with the GradientDescent class.

Parameters

problem	The function being minimized
x0	The initial position during our minimization (it will be overwritten with the optimal position)

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