Function decorators

Namespaces
namespace	roboptim::finiteDifferenceGradientPolicies
	Contains finite difference gradients policies.

Classes
class	roboptim::CachedFunction< T >
	Store previous function computation. More...
class	roboptim::BadGradient< T >
	Exception thrown when a gradient check fails. More...
class	roboptim::BadJacobian< T >
	Exception thrown when a Jacobian check fails. More...
class	roboptim::GenericFiniteDifferenceGradient< T, FdgPolicy >
	Compute automatically a gradient with finite differences. More...

Functions
template<typename T >
std::ostream &	roboptim::operator<< (std::ostream &o, const BadGradient< T > &f)
	Override operator<< to handle exception display. More...
template<typename T >
std::ostream &	roboptim::operator<< (std::ostream &o, const BadJacobian< T > &f)
	Override operator<< to handle exception display. More...
template<typename T >
bool	roboptim::checkGradient (const GenericDifferentiableFunction< T > &function, typename GenericDifferentiableFunction< T >::size_type functionId, typename GenericDifferentiableFunction< T >::const_argument_ref x, typename GenericDifferentiableFunction< T >::value_type threshold=finiteDifferenceThreshold, typename GenericDifferentiableFunction< T >::value_type fd_eps=finiteDifferenceEpsilon)
	Check if a gradient is valid. More...
template<typename T >
void	roboptim::checkGradientAndThrow (const GenericDifferentiableFunction< T > &function, typename GenericDifferentiableFunction< T >::size_type functionId, typename GenericDifferentiableFunction< T >::const_argument_ref x, typename GenericDifferentiableFunction< T >::value_type threshold=finiteDifferenceThreshold, typename GenericDifferentiableFunction< T >::value_type fd_eps=finiteDifferenceEpsilon) throw (BadGradient<T>)
template<typename T >
bool	roboptim::checkJacobian (const GenericDifferentiableFunction< T > &function, typename GenericDifferentiableFunction< T >::const_argument_ref x, typename GenericDifferentiableFunction< T >::value_type threshold=finiteDifferenceThreshold, typename GenericDifferentiableFunction< T >::value_type fd_eps=finiteDifferenceEpsilon)
	Check if a Jacobian is valid. More...
template<typename T >
void	roboptim::checkJacobianAndThrow (const GenericDifferentiableFunction< T > &function, typename GenericDifferentiableFunction< T >::const_argument_ref x, typename GenericDifferentiableFunction< T >::value_type threshold=finiteDifferenceThreshold, typename GenericDifferentiableFunction< T >::value_type fd_eps=finiteDifferenceEpsilon) throw (BadJacobian<T>)

Variables
static const double	roboptim::finiteDifferenceThreshold = 1e-4
	Default threshold for checkGradient. More...
static const double	roboptim::finiteDifferenceEpsilon = 1e-8
	Default epsilon for finite difference class. More...

Detailed Description

Function Documentation

template<typename T >

bool roboptim::checkGradient	(	const GenericDifferentiableFunction< T > &	function,
		typename GenericDifferentiableFunction< T >::size_type	functionId,
		typename GenericDifferentiableFunction< T >::const_argument_ref	x,
		typename GenericDifferentiableFunction< T >::value_type	threshold = finiteDifferenceThreshold,
		typename GenericDifferentiableFunction< T >::value_type	fd_eps = finiteDifferenceEpsilon
	)

Check if a gradient is valid.

Check if a gradient is valid by comparing the distance between its gradient and an automatically computed finite differences gradient.

Parameters

function	function that will be checked
functionId	function id in split representation
x	point where the gradient will be evaluated
threshold	maximum tolerated error

Returns

true if valid, false if not

Examples:

numeric-quadratic-function.cc.

References roboptim::allclose(), roboptim::GenericDifferentiableFunction< T >::gradient(), ROBOPTIM_ALLOW_DEPRECATED_OFF, and ROBOPTIM_ALLOW_DEPRECATED_ON.

template<typename T >

void roboptim::checkGradientAndThrow	(	const GenericDifferentiableFunction< T > &	function,
		typename GenericDifferentiableFunction< T >::size_type	functionId,
		typename GenericDifferentiableFunction< T >::const_argument_ref	x,
		typename GenericDifferentiableFunction< T >::value_type	threshold = finiteDifferenceThreshold,
		typename GenericDifferentiableFunction< T >::value_type	fd_eps = finiteDifferenceEpsilon
	)		throw (BadGradient<T>)

References roboptim::allclose(), roboptim::GenericDifferentiableFunction< T >::gradient(), ROBOPTIM_ALLOW_DEPRECATED_OFF, and ROBOPTIM_ALLOW_DEPRECATED_ON.

template<typename T >

bool roboptim::checkJacobian	(	const GenericDifferentiableFunction< T > &	function,
		typename GenericDifferentiableFunction< T >::const_argument_ref	x,
		typename GenericDifferentiableFunction< T >::value_type	threshold = finiteDifferenceThreshold,
		typename GenericDifferentiableFunction< T >::value_type	fd_eps = finiteDifferenceEpsilon
	)

Check if a Jacobian is valid.

Check if a Jacobian is valid by comparing the distance between the matrix and an automatically computed finite differences Jacobian.

Parameters

function	function that will be checked
x	point where the Jacobian will be evaluated
threshold	maximum tolerated error

Returns

true if valid, false if not

Examples:

numeric-quadratic-function.cc.

References roboptim::allclose(), roboptim::GenericDifferentiableFunction< T >::jacobian(), ROBOPTIM_ALLOW_DEPRECATED_OFF, and ROBOPTIM_ALLOW_DEPRECATED_ON.

template<typename T >

void roboptim::checkJacobianAndThrow	(	const GenericDifferentiableFunction< T > &	function,
		typename GenericDifferentiableFunction< T >::const_argument_ref	x,
		typename GenericDifferentiableFunction< T >::value_type	threshold = finiteDifferenceThreshold,
		typename GenericDifferentiableFunction< T >::value_type	fd_eps = finiteDifferenceEpsilon
	)		throw (BadJacobian<T>)

References roboptim::allclose(), roboptim::GenericDifferentiableFunction< T >::jacobian(), ROBOPTIM_ALLOW_DEPRECATED_OFF, and ROBOPTIM_ALLOW_DEPRECATED_ON.

template<typename T >

std::ostream & roboptim::operator<<	(	std::ostream &	o,
		const BadGradient< T > &	f
	)

Override operator<< to handle exception display.

Parameters

o	output stream used for display
f	function to be displayed

Returns

output stream

template<typename T >

std::ostream & roboptim::operator<<	(	std::ostream &	o,
		const BadJacobian< T > &	f
	)

Override operator<< to handle exception display.

Parameters

o	output stream used for display
f	function to be displayed

Returns

output stream

Variable Documentation

const double roboptim::finiteDifferenceEpsilon = 1e-8

static

Default epsilon for finite difference class.

const double roboptim::finiteDifferenceThreshold = 1e-4

static

Default threshold for checkGradient.