FD::FFNet Class Reference

#include <FFNet.h>

Inheritance diagram for FD::FFNet:

[legend]Collaboration diagram for FD::FFNet:

[legend]List of all members.

Public Member Functions
	FFNet (const Vector< int > &_topo, const Vector< std::string > &functions)
	FFNet (FFNet &net)
	FFNet (const Vector< int > &_topo, const Vector< std::string > &functions, std::vector< float * > &tin, std::vector< float * > &tout)
void	init (const Vector< std::string > &functions)
void	setupLayersAfterRead ()
float *	calc (const float *input, float *value, float *deriv=NULL)
void	learn (float *input, float *output, double *gradient, double *err=NULL, float *calc_output=NULL)
void	calcGradient (std::vector< float * > &tin, std::vector< float * > &tout, Array< float > weights, Array< double > &gradient, double &err)
void	weightedLearn (float *input, float *output, float *learnWeights, double *gradient, double *err=NULL, float *calc_output=NULL)
void	weightedCalcGradient (std::vector< float * > &tin, std::vector< float * > &tout, std::vector< float * > &learnWeights, Array< float > weights, Array< double > &gradient, double &err)
float	totalError (std::vector< float * > tin, std::vector< float * > tout)
int	getNbWeights ()
int	getNbNeurons ()
const Vector< int > &	getTopo ()
const Vector< RCPtr< FFLayer > > &	getLayers ()
const float *	getWeights ()
void	setWeights (float *ptr)
void	setDerivOffset (float d)
void	printOn (std::ostream &out) const
void	readFrom (std::istream &in)
Protected Attributes
Vector< int >	topo
Vector< RCPtr< FFLayer > >	layers
float *	weights
int	nbNeurons
int	nbWeights

---

Detailed Description

Feed-forward neural network (MLP) class

Author:

: Jean-Marc Valin

---

Member Function Documentation

float* FD::FFNet::calc (  const float *  input, float *  value, float *  deriv = NULL )  [inline]

Calculates the network result for a certain input

void FD::FFNet::calcGradient (  std::vector< float * > &  tin, std::vector< float * > &  tout, Array< float >  weights, Array< double > &  gradient, double &  err )

Calculates the gradient for a set of sample

void FD::FFNet::learn (  float *  input, float *  output, double *  gradient, double *  err = NULL, float *  calc_output = NULL )

Calculates the gradient for a single sample

void FD::FFNet::printOn (  std::ostream &  out  )  const [virtual]

Writes an MLP to a stream Implements FD::Object.

void FD::FFNet::readFrom (  std::istream &  in  )  [virtual]

Reads an MLP from a stream Reimplemented from FD::Object.

float FD::FFNet::totalError (  std::vector< float * >  tin, std::vector< float * >  tout )

Mean square error for a complete set

void FD::FFNet::weightedCalcGradient (  std::vector< float * > &  tin, std::vector< float * > &  tout, std::vector< float * > &  learnWeights, Array< float >  weights, Array< double > &  gradient, double &  err )

Calculates the gradient for a set of sample using variable weight

void FD::FFNet::weightedLearn (  float *  input, float *  output, float *  learnWeights, double *  gradient, double *  err = NULL, float *  calc_output = NULL )

Calculates the gradient for a single sample using variable weight

---

Member Data Documentation

Vector<RCPtr<FFLayer> > FD::FFNet::layers [protected]

Pointers to all MLP layers

int FD::FFNet::nbNeurons [protected]

Total number of neurons

int FD::FFNet::nbWeights [protected]

Total number of weights

Vector<int> FD::FFNet::topo [protected]

Network topology

float* FD::FFNet::weights [protected]

Complete weight vector

---

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

• FFNet.h

---