dlr::optimization::GradientFunction< Functor > Class Template Reference

The GradientFunction class template is derived from std::unary_function, and adds one additional member function for computing the gradient of the function output with respect to the argument. More...

#include <gradientFunction.h>

Collaboration diagram for dlr::optimization::GradientFunction< Functor >:

[legend]List of all members.

Public Member Functions
	GradientFunction (const Functor &functor, double epsilon=1.0e-6)
	This is the default constructor.
virtual	~GradientFunction ()
	Destructor.
Functor::argument_type	gradient (const typename Functor::argument_type &theta)
	This method numerically approximates the gradient of this->operator() by divided differences.
Functor::result_type	operator() (const typename Functor::argument_type &theta)
	This method evaluates the function value a the specified point.

---

Detailed Description

template<class Functor>
class dlr::optimization::GradientFunction< Functor >

The GradientFunction class template is derived from std::unary_function, and adds one additional member function for computing the gradient of the function output with respect to the argument.

Template argument Functor is assumed to be a subclass of std::unary_function. Functor::result_type is assumed to be a continuous scalar type. Functor::argument_type is assumed to be a vector or 1D array type which supports the following interface:

argument_type(size_t N): construct an N-element vector. size_t size(): return the number of elements in the vector. element_type& operator[](size_t i): return a reference to the (i)th element of the array. const element_type& operator[](size_t i) const: return a const reference to the (i)th element of the array.

It is further assumed that element type of argument_type is a continuous scalar, and can be implicitly cast to and from double.

Definition at line 49 of file gradientFunction.h.

---

Constructor & Destructor Documentation

template<class Functor>

dlr::optimization::GradientFunction< Functor >::GradientFunction	(	const Functor &	functor,
		double	epsilon = 1.0e-6
	)			[inline]

This is the default constructor.

Parameters:

	functor	This argument is the function object to be adapted.
	epsilon	If the gradient() method is not overridden in a subclass, the gradient will be computed by using symmetric divided differences with a total step size of 2 * epsilon.

Definition at line 64 of file gradientFunction.h.

References DLR_THROW3.

template<class Functor>

virtual dlr::optimization::GradientFunction< Functor >::~GradientFunction

(

)

[inline, virtual]

Destructor.

Definition at line 75 of file gradientFunction.h.

---

Member Function Documentation

template<class Functor>

Functor::argument_type dlr::optimization::GradientFunction< Functor >::gradient

(

const typename Functor::argument_type &

theta

)

[inline]

This method numerically approximates the gradient of this->operator() by divided differences.

This method should often be overridden by a subclass.

Parameters:

theta

The point around which to compute the gradient.

Returns:

The computed gradient.

Definition at line 133 of file gradientFunction.h.

template<class Functor>

Functor::result_type dlr::optimization::GradientFunction< Functor >::operator()

(

const typename Functor::argument_type &

theta

)

[inline]

This method evaluates the function value a the specified point.

Parameters:

theta

The point at which to evaluate the function.

Returns:

The function value at theta.

Definition at line 95 of file gradientFunction.h.

---

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

• gradientFunction.h

---