OpenNN::SolutionsError Class Reference

#include <solutions_error.h>

Inheritance diagram for OpenNN::SolutionsError:

Public Types
enum	SolutionsErrorMethod { SolutionsErrorSum, SolutionsErrorIntegral }

Public Member Functions
	SolutionsError (void)
	SolutionsError (NeuralNetwork *)
	SolutionsError (MathematicalModel *)
	SolutionsError (NeuralNetwork *, MathematicalModel *)
	SolutionsError (const tinyxml2::XMLDocument &)
	SolutionsError (const SolutionsError &)
SolutionsError &	operator= (const SolutionsError &)
bool	operator== (const SolutionsError &) const
virtual	~SolutionsError (void)
const SolutionsErrorMethod &	get_solutions_error_method (void) const
std::string	write_solutions_error_method (void) const
const Vector< double > &	get_solutions_errors_weights (void) const
const double &	get_solution_error_weight (const size_t &) const
void	set_solutions_error_method (const SolutionsErrorMethod &)
void	set_solutions_error_method (const std::string &)
void	set_solutions_errors_weights (const Vector< double > &)
void	set_solution_error_weight (const size_t &, const double &)
void	set_default (void)
virtual Matrix< double >	calculate_target_dependent_variables (const Matrix< double > &) const
void	check (void) const
double	calculate_solutions_error_sum (void) const
double	calculate_solutions_error_integral (void) const
virtual double	calculate_performance (void) const
virtual double	calculate_performance (const Vector< double > &) const
std::string	write_performance_term_type (void) const
std::string	write_information (void) const
virtual void	print (void) const
tinyxml2::XMLDocument *	to_XML (void) const
void	from_XML (const tinyxml2::XMLDocument &)
Public Member Functions inherited from OpenNN::PerformanceTerm
	PerformanceTerm (void)
	PerformanceTerm (NeuralNetwork *)
	PerformanceTerm (DataSet *)
	PerformanceTerm (MathematicalModel *)
	PerformanceTerm (NeuralNetwork *, DataSet *)
	PerformanceTerm (NeuralNetwork *, MathematicalModel *)
	PerformanceTerm (NeuralNetwork *, MathematicalModel *, DataSet *)
	PerformanceTerm (const tinyxml2::XMLDocument &)
	PerformanceTerm (const PerformanceTerm &)
virtual	~PerformanceTerm (void)
virtual PerformanceTerm &	operator= (const PerformanceTerm &)
virtual bool	operator== (const PerformanceTerm &) const
NeuralNetwork *	get_neural_network_pointer (void) const
MathematicalModel *	get_mathemtaical_model_pointer (void) const
DataSet *	get_data_set_pointer (void) const
NumericalDifferentiation *	get_numerical_differentiation_pointer (void) const
const bool &	get_display (void) const
bool	has_neural_network (void) const
bool	has_mathematical_model (void) const
bool	has_data_set (void) const
bool	has_numerical_differentiation (void) const
virtual void	set (void)
virtual void	set (NeuralNetwork *)
virtual void	set (DataSet *)
virtual void	set (MathematicalModel *)
virtual void	set (NeuralNetwork *, DataSet *)
virtual void	set (NeuralNetwork *, MathematicalModel *)
virtual void	set (NeuralNetwork *, MathematicalModel *, DataSet *)
void	set (const PerformanceTerm &)
virtual void	set_neural_network_pointer (NeuralNetwork *)
virtual void	set_mathematical_model_pointer (MathematicalModel *)
virtual void	set_data_set_pointer (DataSet *)
void	set_numerical_differentiation_pointer (NumericalDifferentiation *)
void	set_display (const bool &)
void	construct_numerical_differentiation (void)
void	delete_numerical_differentiation_pointer (void)
Vector< Vector< double > >	calculate_layers_delta (const Vector< Vector< double > > &, const Vector< double > &) const
Vector< Vector< double > >	calculate_layers_delta (const Vector< Vector< double > > &, const Vector< double > &, const Vector< double > &) const
Matrix< Matrix< double > >	calculate_interlayers_Delta (const Vector< Vector< double > > &, const Vector< Vector< double > > &, const Matrix< Matrix< double > > &, const Vector< double > &, const Matrix< double > &, const Vector< Vector< double > > &) const
Vector< double >	calculate_point_gradient (const Vector< double > &, const Vector< Vector< double > > &, const Vector< Vector< double > > &) const
Vector< double >	calculate_point_gradient (const Vector< Matrix< double > > &, const Vector< Vector< double > > &) const
Matrix< double >	calculate_point_Hessian (const Vector< Vector< double > > &, const Vector< Vector< Vector< double > > > &, const Matrix< Matrix< double > > &, const Vector< Vector< double > > &, const Matrix< Matrix< double > > &) const
virtual double	calculate_generalization_performance (void) const
virtual Vector< double >	calculate_gradient (void) const
virtual Vector< double >	calculate_gradient (const Vector< double > &) const
virtual Matrix< double >	calculate_Hessian (void) const
virtual Matrix< double >	calculate_Hessian (const Vector< double > &) const
virtual Vector< double >	calculate_terms (void) const
virtual Vector< double >	calculate_terms (const Vector< double > &) const
virtual Matrix< double >	calculate_terms_Jacobian (void) const
virtual PerformanceTerm::FirstOrderTerms	calculate_first_order_terms (void) const
virtual std::string	to_string (void) const
size_t	calculate_Kronecker_delta (const size_t &, const size_t &) const

Protected Attributes
NumericalIntegration	numerical_integration
SolutionsErrorMethod	solutions_error_method
Vector< double >	solutions_errors_weights
Protected Attributes inherited from OpenNN::PerformanceTerm
NeuralNetwork *	neural_network_pointer
DataSet *	data_set_pointer
MathematicalModel *	mathematical_model_pointer
NumericalDifferentiation *	numerical_differentiation_pointer
bool	display

Detailed Description

This class implements a performance term called solutions error. This performance term is measured on a mathematical model. It computes the difference between the solution from a mathematical model and a given target solution. This type of functional is very used in optimal control problems, in which the neural network represents the control variables. The first and second derivatives of this type of performance term are calculated with numerical differentiation.

Definition at line 45 of file solutions_error.h.

Constructor & Destructor Documentation

OpenNN::SolutionsError::SolutionsError ( void  )

explicit

Default constructor. It creates a solutions error performance term not associated to any neural network and not measured on any mathematical model. It also initializes all the rest of class members to their default values.

Definition at line 28 of file solutions_error.cpp.

OpenNN::SolutionsError::SolutionsError ( NeuralNetwork *  new_neural_network_pointer )

explicit

Neural network constructor. It creates a solutions error associated to a neural network, but not to a . It also initializes all the rest of class members to their default values.

Parameters

new_neural_network_pointer

Pointer to a neural network object.

Definition at line 43 of file solutions_error.cpp.

OpenNN::SolutionsError::SolutionsError ( MathematicalModel *  new_mathematical_model_pointer )

explicit

Mathematical model constructor. It creates solutions error object not associated to any neural network but to be measured on a given mathematical model object. It also initializes all the rest of class members to their default values.

Parameters

new_mathematical_model_pointer

Pointer to a mathematical model object.

Definition at line 59 of file solutions_error.cpp.

OpenNN::SolutionsError::SolutionsError ( NeuralNetwork *  new_neural_network_pointer, MathematicalModel *  new_mathematical_model_pointer  )

explicit

Neural network and mathematical model constructor. It creates a solutions error functional associated to a neural network and measured on a mathematical model. It also initializes all the rest of class members to their default values.

Parameters

new_neural_network_pointer	Pointer to a neural network object.
new_mathematical_model_pointer	Pointer to a mathematical model object.

Definition at line 76 of file solutions_error.cpp.

OpenNN::SolutionsError::SolutionsError ( const tinyxml2::XMLDocument &  solutions_error_document )

explicit

xml constructor. It creates a solutions error performance term not associated to any neural network and not measured on any mathematical model. It also initializes all the rest of class members with values taken from a TinyXML document->

Parameters

solutions_error_document

XML document containing the class members.

Definition at line 92 of file solutions_error.cpp.

OpenNN::SolutionsError::SolutionsError

(

const SolutionsError &

other_solutions_error

)

Copy constructor. It creates a copy of an existing solutions error object.

Parameters

other_solutions_error

Solutions error object to be copied.

Definition at line 107 of file solutions_error.cpp.

OpenNN::SolutionsError::~SolutionsError ( void  )

virtual

Destructor. This destructor does not delete any object. The numerical differentiation object is deleted by the performance term destructor.

Definition at line 138 of file solutions_error.cpp.

Member Function Documentation

double OpenNN::SolutionsError::calculate_performance ( const Vector< double > &  parameters ) const

virtual

Calculates the solutions error value of the neural network for a given set of parameters.

Parameters

parameters

Vector of potential neural network parameters.

Implements OpenNN::PerformanceTerm.

Definition at line 583 of file solutions_error.cpp.

double OpenNN::SolutionsError::calculate_solutions_error_integral

(

void

)

const

Todo:

Definition at line 529 of file solutions_error.cpp.

double OpenNN::SolutionsError::calculate_solutions_error_sum

(

void

)

const

Calculates the solutions error performance as the weighted mean squared error between the solutions to the mathematical model and the target solutions.

Definition at line 481 of file solutions_error.cpp.

void OpenNN::SolutionsError::check ( void  ) const

virtual

Checks that:

• The pointer to the neural network is not NULL.
• The pointer to the multilayer perceptron inside the neural network is not NULL.
• The pointer to the mathematical method is not NULL.
• The size of the solutions errors weights is equal to the number of dependent variables.

If some of the conditions above is not hold, an exception is thrown.

Reimplemented from OpenNN::PerformanceTerm.

Definition at line 401 of file solutions_error.cpp.

void OpenNN::SolutionsError::from_XML ( const tinyxml2::XMLDocument &  document )

virtual

Loads a solutions error object from a XML document.

Parameters

document

TinyXML document with the members of the solutions error object.

Reimplemented from OpenNN::PerformanceTerm.

Definition at line 718 of file solutions_error.cpp.

SolutionsError & OpenNN::SolutionsError::operator=

(

const SolutionsError &

other_solutions_error

)

Assignment operator. It assigns to this solutions error object the members from another solutions error object.

Parameters

other_solutions_error

Solutions error object to be copied.

Definition at line 151 of file solutions_error.cpp.

bool OpenNN::SolutionsError::operator==

(

const SolutionsError &

other_solutions_error

)

const

Equal to operator. It compares this object to another object, and returns true if both have the same member data, and false otherwise.

Definition at line 176 of file solutions_error.cpp.

void OpenNN::SolutionsError::set_default ( void  )

virtual

Sets the default values for an object of the solutions error class:

• Solutions error method: Solutions error sum.
• Solutions error weights: 1 for all dependent variables.
• Display: True.

Reimplemented from OpenNN::PerformanceTerm.

Definition at line 347 of file solutions_error.cpp.

void OpenNN::SolutionsError::set_solution_error_weight	(	const size_t &	i,
		const double &	new_solution_error_weight
	)

Sets a new weight for the solution error of a single dependent variable.

Parameters

i	Index of dependent variable.
new_solution_error_weight	Weight for the error between the above dependent variable solution and the corresponding target solution.

Definition at line 332 of file solutions_error.cpp.

void OpenNN::SolutionsError::set_solutions_error_method

(

const SolutionsErrorMethod &

new_solutions_error_method

)

Sets a new method for calculating the error between the solution of a mathematical model and a target solution.

Parameters

new_solutions_error_method

Method for evaluating this performance term.

Definition at line 281 of file solutions_error.cpp.

void OpenNN::SolutionsError::set_solutions_error_method

(

const std::string &

new_solutions_error_method

)

Sets a new solutions error method from a string with that method's name.

Parameters

new_solutions_error_method

String with the name of the method.

Definition at line 292 of file solutions_error.cpp.

void OpenNN::SolutionsError::set_solutions_errors_weights

(

const Vector< double > &

new_solutions_errors_weights

)

Sets new weights for the solution error of every dependent variable.

Parameters

new_solutions_errors_weights

Weights for the errors between the dependent variable solutions and the corresponding target solutions.

Definition at line 320 of file solutions_error.cpp.

std::string OpenNN::SolutionsError::write_information ( void  ) const

virtual

Returns a string with the default information of the performance term. It will be used by the training strategy to monitor the training process. By default this information is empty.

Reimplemented from OpenNN::PerformanceTerm.

Definition at line 618 of file solutions_error.cpp.

---

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

• solutions_error.h
• solutions_error.cpp