train-FFN.C File Reference

---

Detailed Description

train an multilayer feed-forward netwark with backpropagation

Definition in file train-FFN.C.

#include "Component/ModelManager.H"
#include "Gist/FFN.H"
#include "Raster/Raster.H"
#include "Util/MathFunctions.H"
#include "Gist/trainUtils.H"
#include "Image/MatrixOps.H"
#include <vector>

Include dependency graph for train-FFN.C:

Go to the source code of this file.

Defines
#define	ABSOLUTE   0
#define	MIXTURE   1
#define	ERR_THRESHOLD   .01
#define	MAX_EPOCH   1000
Functions
void	setupCases (std::string folder, std::string fname)
void	train ()
void	test ()
void	run (int isTest)
void	diff (Image< double > ideal, Image< double > out, double &tErr, int &tFc, int &tIc)
int	main (const int argc, const char **argv)
Variables
rutz::shared_ptr< FeedForwardNetwork >	ffn
int	nSamples = 0
std::vector< Image< double > >	in
std::vector< Image< double > >	out
Image< double >	pcaIcaMatrix
rutz::shared_ptr< FFNtrainInfo >	info

---

Define Documentation

#define ABSOLUTE   0

training a neural network Definition at line 138 of file train-FFN.C. Referenced by setupCases().

---