graphite::optimizer Namespace Reference

Optimization algorithms for Graphite. More...

Classes
class	AdamOptions
class	GradientDescentOptions
class	LevenbergMarquardtOptions
	Levenberg-Marquardt options. More...

Functions
template<typename T , typename S >
bool	adam (Graph< T, S > *graph, AdamOptions< T, S > *options)
	Adam optimization algorithm.
template<typename T , typename S >
bool	gradient_descent (Graph< T, S > *graph, GradientDescentOptions< T, S > *options)
	Naive gradient descent optimization algorithm.
template<typename T , typename S >
T	compute_rho (Graph< T, S > *graph, thrust::device_vector< T > &delta_x, const T chi2, const T new_chi2, const T mu, const bool step_is_good)
template<typename T , typename S >
bool	levenberg_marquardt (Graph< T, S > *graph, LevenbergMarquardtOptions< T, S > *options)
	Levenberg-Marquardt optimization algorithm.
template<typename T , typename S >
bool	levenberg_marquardt2 (Graph< T, S > *graph, LevenbergMarquardtOptions< T, S > *options)
	Levenberg-Marquardt with similar early termination stopping criteria to ORB-SLAM.

Detailed Description

Optimization algorithms for Graphite.

Function Documentation

adam()

template<typename T , typename S >

bool graphite::optimizer::adam	(	Graph< T, S > *	graph,
		AdamOptions< T, S > *	options
	)

Adam optimization algorithm.

Template Parameters

T	Scalar type
S	Scalar type

Parameters

graph	Graph to optimize
options	Options for the optimization

Returns

true if optimization completed successfully, false otherwise

gradient_descent()

template<typename T , typename S >

bool graphite::optimizer::gradient_descent	(	Graph< T, S > *	graph,
		GradientDescentOptions< T, S > *	options
	)

Naive gradient descent optimization algorithm.

Template Parameters

T	Scalar type
S	Scalar type

Parameters

graph	Graph to optimize
options	Options for the optimization

Returns

true if optimization completed successfully, false otherwise

levenberg_marquardt()

template<typename T , typename S >

bool graphite::optimizer::levenberg_marquardt	(	Graph< T, S > *	graph,
		LevenbergMarquardtOptions< T, S > *	options
	)

Levenberg-Marquardt optimization algorithm.

Template Parameters

T	Scalar type
S	Scalar type

Parameters

graph	Graph to optimize
options	Options for the optimization

Returns

true if optimization completed successfully, false otherwise

levenberg_marquardt2()

template<typename T , typename S >

bool graphite::optimizer::levenberg_marquardt2	(	Graph< T, S > *	graph,
		LevenbergMarquardtOptions< T, S > *	options
	)

Levenberg-Marquardt with similar early termination stopping criteria to ORB-SLAM.

Template Parameters

T	Scalar type
S	Scalar type

Parameters

graph	Graph to optimize
options	Options for the optimization

Returns

true if optimization completed successfully, false otherwise