GmpFemNeuralNetwork Class Reference

A simple class to store additional Data required by Fem physics. More...

#include <gmpFemNeuralNetwork.h>

Collaboration diagram for GmpFemNeuralNetwork:

Public Member Functions
	GmpFemNeuralNetwork (const GmLogCategory &logger)
	Constructor for an Artificial Neural Network used by physics and materials. More...
virtual	~GmpFemNeuralNetwork ()
	Destructor.
bool	loadNetArchitecture (LuaTable &table)
	Loads the neural network architecture. More...
bool	loadNetWeights (LuaTable &table)
	Loads the neural network weights. More...
bool	loadNetBias (LuaTable &table)
	Loads the neural network bias. More...
bool	loadNetInputSettings (LuaTable &table)
	Loads input settings of the neural network.
bool	loadNetOutputSettings (LuaTable &table)
	Loads output settings of the neural network.
GmMatrix *	readNetWeight (int i)
	reads weight at position i
bool	predict (GmVector &x, GmVector &y)
	Evaluates the neural network. More...
bool	normalization (GmVector &x, GmVector &xn)
	Normalization of input variables. More...
bool	inverseNormalization (GmVector &x, GmVector &xn)
	Inverse normalization of output variables.
GmVector	tanh (GmVector &x)
	Evaluates the hyperbolic tangent sigmoid activation function. More...
GmVector	logsig (GmVector &x)
	Evaluates the logistic sigmoid activation function. More...
int	stringOptionIndex (QString value, const char **optionNames)
	Looks for value in optionNames. If found returns the index.

Private Member Functions
const char **	activationFcnOptions ()
	Returns a NULL terminated vector with the available options for the solver type.

Private Attributes
GmVector	_netConf
	Vector to define the ANN architecture.
GmMatrix	_inputsettings
	Matrix with {min, max} of input variables (input range)
GmMatrix	_outputSettings
	Matrix with {min, max} of output variables (output range)
QList< QString >	_activationFcn
	List with activation functions.
QList< GmVector >	_netBias
	List with neural bias.
QList< GmMatrix >	_netWeights
	List with neural weights.
const GmLogCategory &	_logger
	Reference to the plugin private logger.

Detailed Description

A simple class to store additional Data required by Fem physics.

Constructor & Destructor Documentation

GmpFemNeuralNetwork()

GmpFemNeuralNetwork::GmpFemNeuralNetwork

(

const GmLogCategory &

logger

)

Constructor for an Artificial Neural Network used by physics and materials.

IMPORTANT: The ANN takes ownership of the given data vector and will destroy it on its destructor.

Member Function Documentation

loadNetArchitecture()

bool GmpFemNeuralNetwork::loadNetArchitecture

(

LuaTable &

table

)

Loads the neural network architecture.

Loads neural network architecture.

loadNetBias()

bool GmpFemNeuralNetwork::loadNetBias

(

LuaTable &

table

)

Loads the neural network bias.

Loads neural network bias.

loadNetWeights()

bool GmpFemNeuralNetwork::loadNetWeights

(

LuaTable &

table

)

Loads the neural network weights.

Loads neural network weights.

logsig()

GmVector GmpFemNeuralNetwork::logsig

(

GmVector &

x

)

Evaluates the logistic sigmoid activation function.

Evaluates the logistic sigmoid activation function y = 1.0 / (1.0 + exp(-x))

normalization()

bool GmpFemNeuralNetwork::normalization	(	GmVector &	x,
		GmVector &	xn
	)

Normalization of input variables.

Normaization of input variables.

predict()

bool GmpFemNeuralNetwork::predict	(	GmVector &	x,
		GmVector &	y
	)

Evaluates the neural network.

Loads output settings of the neural network.

tanh()

GmVector GmpFemNeuralNetwork::tanh

(

GmVector &

x

)

Evaluates the hyperbolic tangent sigmoid activation function.

Evaluates the hyperbolic tangent sigmoid activation function y = tanh(x) = 2.0 /(1.0 + exp(-2.0 * x)) - 1.0.

---

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

• inc/gmpFemNeuralNetwork.h
• src/gmpFemNeuralNetwork.cpp