#include <kerdensom.h>

Inheritance diagram for KerDenSOM:

Collaboration diagram for KerDenSOM:

Public Member Functions
	KerDenSOM (double _reg0, double _reg1, unsigned long _annSteps, double _epsilon, unsigned long _nSteps)

virtual	~KerDenSOM ()

void	nSteps (const unsigned long &_nSteps)

void	setAnnSteps (const unsigned long &_annSteps)

virtual void	train (FuzzyMap &_som, TS &_examples, FileName &_fn_vectors, bool _update=false, double _sigma=0, bool _saveIntermediate=false)=0

virtual double	test (const FuzzyMap &_som, const TS &_examples) const

virtual double	functional (const TS _examples, const FuzzyMap _som, double _sigma, double _reg, double &_likelihood, double &_penalty)=0

Public Member Functions inherited from ClassificationAlgorithm< FuzzyMap >
	ClassificationAlgorithm (const std::string &_ID="")

virtual	~ClassificationAlgorithm ()

virtual void	train (DS &_ds, const TS &_examples) const

virtual void	train (DS &_ds, TS &_examples) const

virtual double	test (const DS &_ds, const TS &_examples) const=0

virtual void	printSelf (std::ostream &_os) const

virtual const std::string &	setID () const

virtual std::string &	setID ()

void	setListener (BaseListener *_listener)

Protected Member Functions
virtual void	train (FuzzyMap &_som, const TS &_examples) const

virtual double	updateU (FuzzyMap _som, const TS _examples, const double &_sigma, double &_alpha)=0

virtual double	updateSigmaII (FuzzyMap _som, const TS _examples, const double &_reg, const double &_alpha)=0

virtual double	codeDens (const FuzzyMap _som, const FeatureVector _example, double _sigma) const =0

virtual void	updateV (FuzzyMap _som, const TS _examples, const double &_sigma)

virtual double	mainIterations (FuzzyMap _som, const TS _examples, double &_sigma, const double &_reg)

virtual void	initU (FuzzyMap *_som)

virtual void	updateV1 (FuzzyMap _som, const TS _examples)

virtual void	updateU1 (FuzzyMap _som, const TS _examples)

virtual double	updateSigmaI (FuzzyMap _som, const TS _examples)

void	showX (const TS *_ts)

void	showV (FuzzyMap *_som)

void	showU (FuzzyMap _som, const TS _ts)

Protected Attributes
double	sigma

size_t	annSteps

double	reg0

double	reg1

double	epsilon

size_t	somNSteps

size_t	numNeurons

size_t	numVectors

size_t	dim

std::vector< std::vector< double > >	tmpMap

std::vector< double >	tmpD

std::vector< double >	tmpD1

std::vector< double >	tmpDens

std::vector< double >	tmpV

Protected Attributes inherited from ClassificationAlgorithm< FuzzyMap >
std::string	ID

BaseListener *	listener

Additional Inherited Members
Public Types inherited from ClassificationAlgorithm< FuzzyMap >
typedef FuzzyMap	DS

typedef DS::TS	TS

Detailed Description

This class trains a Smoothly Distributed Kernel Probability Density Estimator Self Organizing Map

Definition at line 44 of file kerdensom.h.

Constructor & Destructor Documentation

◆ KerDenSOM()

KerDenSOM::KerDenSOM	(	double	_reg0,
		double	_reg1,
		unsigned long	_annSteps,
		double	_epsilon,
		unsigned long	_nSteps
	)

inline

Constructs the algorithm Parameter: _reg0 Initial regularization factor Parameter: _reg1 Final regularization factor Parameter: _annSteps Number of steps in deterministic annealing Parameter: _epsilon Stopping criterion Parameter: _nSteps Number of training steps

Definition at line 56 of file kerdensom.h.

             : ClassificationAlgorithm<FuzzyMap>(), annSteps(_annSteps), reg0(_reg0), reg1(_reg1),
             epsilon(_epsilon), somNSteps(_nSteps)
     {};

◆ ~KerDenSOM()

virtual KerDenSOM::~KerDenSOM ( )

inlinevirtual

Virtual destructor

Definition at line 65 of file kerdensom.h.

66 {};

Member Function Documentation

◆ codeDens()

virtual double KerDenSOM::codeDens	(	const FuzzyMap *	_som,
		const FeatureVector *	_example,
		double	_sigma
	)		const

protectedpure virtual

Implemented in GaussianKerDenSOM.

◆ functional()

virtual double KerDenSOM::functional	(	const TS *	_examples,
		const FuzzyMap *	_som,
		double	_sigma,
		double	_reg,
		double &	_likelihood,
		double &	_penalty
	)

pure virtual

Determines the functional value. Returns the likelihood and penalty parts of the functional

◆ initU()

void KerDenSOM::initU ( FuzzyMap * _som )

protectedvirtual

Special Initialization of the Us

Definition at line 334 of file kerdensom.cpp.

 {
     unsigned cc;
     unsigned vv;
 
     // Take random samples
     randomize_random_generator();
     for (vv = 0; vv < numVectors; vv++)
     {
         double t = 0.;
         for (cc = 0; cc < numNeurons; cc++)
         {
             _som->memb[vv][cc] = (floatFeature) rnd_unif();
             t += _som->memb[vv][cc];
         }
         for (cc = 0; cc < numNeurons; cc++)
             _som->memb[vv][cc] /= (floatFeature) t;
     }
 }

◆ mainIterations()

double KerDenSOM::mainIterations	(	FuzzyMap *	_som,
		const TS *	_examples,
		double &	_sigma,
		const double &	_reg
	)

protectedvirtual

Definition at line 173 of file kerdensom.cpp.

 {
     int verbosity = listener->getVerbosity();
     double stopError = 1e10;
     double alpha;
     double ts2;
     size_t iter = 0;
     if (somNSteps == 0)
         return stopError;
     if (verbosity == 1 || verbosity == 3)
         listener->OnInitOperation(somNSteps);
     do
     {
         iter++;
         updateU(_som, _examples, _sigma, alpha);
         ts2 = updateSigmaII(_som, _examples, _reg, alpha);
         stopError = fabs(_sigma / ts2 - 1.);
         _sigma = ts2;
         updateV(_som, _examples, _reg);
         if (verbosity == 1 || verbosity == 3)
             listener->OnProgress(iter);
         if (verbosity >= 2)
         {
             char s[100];
             sprintf(s, "Iteration %d of %d. variation: %g\n", (int)iter, (int)somNSteps, stopError);
             listener->OnReportOperation((std::string) s);
         }
     }
     while ((stopError > epsilon) && (iter < somNSteps));
     if (verbosity == 1 || verbosity == 3)
         listener->OnProgress(somNSteps);
     return stopError;
 }

◆ nSteps()

void KerDenSOM::nSteps ( const unsigned long & _nSteps )

Sets the number of training steps Parameter: _nSteps Number of training steps

Definition at line 40 of file kerdensom.cpp.

 {
     somNSteps = _nSteps;
 }

◆ setAnnSteps()

void KerDenSOM::setAnnSteps ( const unsigned long & _annSteps )

Sets the number of deterministic annealing training steps Parameter: _annSteps Number of steps

Definition at line 61 of file kerdensom.cpp.

 {
     annSteps = _annSteps;
 }

◆ showU()

void KerDenSOM::showU	(	FuzzyMap *	_som,
		const TS *	_ts
	)

protected

Shows the U ("Membership")

Definition at line 462 of file kerdensom.cpp.

 {
     std::cout << " Memberships (1..nd,1..ni,1..nj)" << std::endl;
     for (size_t i = 0; i <  _ts->size(); i++)
     {
         for (size_t j = 0; j < _som->size(); j++)
         {
             int tmpj = _som->indexToPos(j).first;
             int tmpi = _som->indexToPos(j).second;
             std::cout << i + 1 << "  " << tmpi + 1 << "  " << tmpj + 1 << "  " << _som->memb[i][j] << std::endl;
         }
     }
 }

◆ showV()

void KerDenSOM::showV ( FuzzyMap * _som )

protected

Shows the code vectors

Definition at line 445 of file kerdensom.cpp.

 {
     std::cout << "Code vectors (1..ni, 1..nj, 1..nv) "  << std::endl;
     for (size_t i = 0; i < _som->size(); i++)
     {
         int tmpj = _som->indexToPos(i).first;
         int tmpi = _som->indexToPos(i).second;
         for (size_t j = 0; j < _som->theItems[0].size(); j++)
             std::cout << tmpi + 1 << "  " << tmpj + 1 << "  " << j + 1 << "  " << _som->theItems[i][j] << std::endl;
     }
 }

◆ showX()

void KerDenSOM::showX ( const TS * _ts )

protected

Shows the data

Definition at line 427 of file kerdensom.cpp.

 {
     std::cout << "Data (1..nd, 1..nv) "  << std::endl;
     for (size_t i = 0; i < _ts->size(); i++)
     {
         for (size_t j = 0; j < _ts->theItems[0].size(); j++)
         {
             std::cout << i + 1 << "  " << j + 1 << "  " << _ts->theItems[i][j] << std::endl;
         }
     }
 }

◆ test()

double KerDenSOM::test	(	const FuzzyMap &	_som,
		const TS &	_examples
	)		const

virtual

Tests the KerDenSOM Parameter: _som The KerDenSom to test Parameter: _examples The testing set

Tests the KerDenSOM Parameter: _som The KerDenSom to test Parameter: _examples The training set of examples

Definition at line 74 of file kerdensom.cpp.

 {
 
     // Defines verbosity level
     int verbosity = listener->getVerbosity();
     if (verbosity)
     {
         listener->OnReportOperation((std::string) "\nEstimating quantization error....\n");
         listener->OnInitOperation(_examples.size());
     }
 
 
     /* Scan all data entries */
     double qerror = 0.0;
     for (size_t i = 0; i < _examples.size(); i++)
     {
         SomIn& theBest = _som.fuzzyTest(i); // get the best
         qerror += euclideanDistance(theBest, _examples.theItems[i]);
         if (verbosity)
         {
             auto tmp = (int)((_examples.size() * 5) / 100);
             if ((tmp == 0) && (i != 0))
                 tmp = i;
             else
                 tmp = 1;
             if ((i % tmp) == 0)
                 listener->OnProgress(i);
         }
     }
     if (verbosity)
         listener->OnProgress(_examples.size());
     return (qerror / (double) _examples.size());
 
 }

◆ train() [1/2]

virtual void KerDenSOM::train	(	FuzzyMap &	_som,
		TS &	_examples,
		FileName &	_fn_vectors,
		bool	_update = `false`,
		double	_sigma = `0`,
		bool	_saveIntermediate = `false`
	)

pure virtual

Trains the KerDenSOM Parameter: _som The KerDenSom to train Parameter: _ts The training set Parameter: _update True if uses _som as starting point for training. Parameter: _sigma If update = true, uses this sigma for the training.

◆ train() [2/2]

virtual void KerDenSOM::train	(	FuzzyMap &	_som,
		const TS &	_examples
	)		const

inlineprotectedvirtual

Declaration of virtual method

Definition at line 144 of file kerdensom.h.

145 {};

◆ updateSigmaI()

double KerDenSOM::updateSigmaI	(	FuzzyMap *	_som,
		const TS *	_examples
	)

protectedvirtual

Definition at line 211 of file kerdensom.cpp.

 {
     double t = 0;
 
     // Computing Sigma (Part I)
     for (size_t vv = 0; vv < numVectors; vv++)
     {
         const FeatureVector &example=_examples->theItems[vv];
         const floatFeature *ptrExample0=&example[0];
         for (size_t cc = 0; cc < numNeurons; cc++)
         {
             double r = 0.0;
             const floatFeature *ptrExample=ptrExample0;
             const floatFeature *ptrCodeVector=&(_som->theItems[cc][0]);
             for (size_t j = 0; j < dim; j++)
             {
                 double diff=(double)(*ptrExample++) - (double)(*ptrCodeVector++);
                 r += diff*diff;
             }
             t += r * (double)(_som->memb[vv][cc]);
         }
     }
     return (double)(t / (double)(numVectors*dim));
 }

◆ updateSigmaII()

virtual double KerDenSOM::updateSigmaII	(	FuzzyMap *	_som,
		const TS *	_examples,
		const double &	_reg,
		const double &	_alpha
	)

protectedpure virtual

◆ updateU()

virtual double KerDenSOM::updateU	(	FuzzyMap *	_som,
		const TS *	_examples,
		const double &	_sigma,
		double &	_alpha
	)

protectedpure virtual

◆ updateU1()

void KerDenSOM::updateU1	(	FuzzyMap *	_som,
		const TS *	_examples
	)

protectedvirtual

Special Initialization of Membership Matrix (Fuzzy c-means style)

Definition at line 287 of file kerdensom.cpp.

 {
 
     double auxProd;
     double auxDist;
     double tmp;
     size_t k;
     size_t j;
     size_t i;
 
     // Update Membership matrix
     for (k = 0; k < numVectors; k++)
     {
         auxProd = 1;
         for (j = 0; j < numNeurons; j++)
             auxProd *= euclideanDistance(_som->theItems[j], _examples->theItems[k]);
 
         if (auxProd == 0.)
         { // Apply k-means criterion (Data-CB) must be > 0
             for (j = 0; j < numNeurons; j ++)
                 if (euclideanDistance(_som->theItems[j], _examples->theItems[k]) == 0.)
                     _som->memb[k][j] = 1.0;
                 else
                     _som->memb[k][j] =  0.0;
         }
         else
         {
             for (i = 0; i < numNeurons; i ++)
             {
                 auxDist = 0;
                 for (j = 0; j < numNeurons; j ++)
                 {
                     tmp = euclideanDistance(_som->theItems[i], _examples->theItems[k]) /
                             euclideanDistance(_som->theItems[j], _examples->theItems[k]);
                     auxDist += pow(tmp, 2);
                 } // for j
                 _som->memb[k][i] = (floatFeature) 1.0 / auxDist;
             } // for i
         } // if auxProd
     } // for k
 }

◆ updateV()

void KerDenSOM::updateV	(	FuzzyMap *	_som,
		const TS *	_examples,
		const double &	_reg
	)

protectedvirtual

Update Code Vectors

Definition at line 116 of file kerdensom.cpp.

 {
     unsigned t2 = 0;  // Iteration index
 
     // Calculate Temporal scratch values
     for (size_t cc = 0; cc < numNeurons; cc++)
     {
         double *ptrTmpMap_cc=&(tmpMap[cc][0]);
         memset(ptrTmpMap_cc,0,dim*sizeof(double));
         double &tmpDens_cc=tmpDens[cc];
         if (_reg != 0)
             tmpDens_cc = _reg * _som->getLayout().numNeig(_som, (SomPos) _som->indexToPos(cc));
         else
             tmpDens_cc = 0.;
         for (size_t vv = 0; vv < numVectors; vv++)
         {
             auto tmpU = (double) _som->memb[vv][cc];
             tmpDens_cc += tmpU;
             const floatFeature * ptrExample=&(_examples->theItems[vv][0]);
             for (size_t j = 0; j < dim; j++)
                 ptrTmpMap_cc[j] +=  tmpU * ptrExample[j];
         }
     }
 
     // Update Code vectors using a sort of Gauss-Seidel iterative algorithm.
     // Usually 100 iterations are enough.
     double convergence = 1;
     double stopError2;
     double stopError1;
     double *ptrTmpV=&(tmpV[0]);
     while ((convergence > 1e-5) && (t2 < 100))
     {
         t2++;
         stopError2 = 0;
         stopError1 = 0;
         for (size_t cc = 0; cc < numNeurons; cc++)
         {
             if (_reg != 0)
                 _som->localAve(_som->indexToPos(cc), tmpV);
             double *ptrTmpMap_cc=&(tmpMap[cc][0]);
             double iTmpDens_cc=1.0/tmpDens[cc];
             floatFeature *ptrCodeVector_cc=&(_som->theItems[cc][0]);
             for (size_t j = 0; j < dim; j++)
             {
                 double tmpU = (ptrTmpMap_cc[j] + ptrTmpV[j] * _reg) * iTmpDens_cc;
                 stopError1 += fabs(ptrCodeVector_cc[j] - tmpU);
                 stopError2 += fabs(tmpU);
                 ptrCodeVector_cc[j] = (floatFeature) tmpU;
             }
         } // for
         convergence = stopError1 / stopError2;
     } // while
 }

◆ updateV1()

void KerDenSOM::updateV1	(	FuzzyMap *	_som,
		const TS *	_examples
	)

protectedvirtual

Special Initialization of Code Vectors

Definition at line 246 of file kerdensom.cpp.

 {
     for (size_t cc = 0; cc < numNeurons; cc++)
     {
         std::vector<double> &tmpMap_cc=tmpMap[cc];
         double *ptrTmpMap_cc0=&tmpMap_cc[0];
         memset(ptrTmpMap_cc0,0,dim*sizeof(double));
         double &tmpDens_cc=tmpDens[cc];
         tmpDens_cc = 0.0;
         for (size_t vv = 0; vv < numVectors; vv++)
         {
             double tmpU = _som->memb[vv][cc];
             tmpDens_cc += tmpU;
             const floatFeature *ptrExample=&(_examples->theItems[vv][0]);
             double *ptrTmpMap_cc=ptrTmpMap_cc0;
             for (size_t j = 0; j < dim; j++)
             {
                 *ptrTmpMap_cc += tmpU * (*ptrExample);
                 ++ptrExample;
                 ++ptrTmpMap_cc;
             }
         }
     }
 
     for (size_t cc = 0; cc < numNeurons; cc++)
     {
         const std::vector<double> &tmpMap_cc=tmpMap[cc];
         double itmpDens_cc=1.0/tmpDens[cc];
         FeatureVector &codevector=_som->theItems[cc];
         for (size_t j = 0; j < dim; j++)
         {
             double tmpU =tmpMap_cc[j] * itmpDens_cc;
             codevector[j] = (floatFeature) tmpU;
         }
     } // for
 }

Member Data Documentation

◆ annSteps

size_t KerDenSOM::annSteps

protected

Definition at line 124 of file kerdensom.h.

◆ dim

size_t KerDenSOM::dim

protected

Definition at line 135 of file kerdensom.h.

◆ epsilon

double KerDenSOM::epsilon

protected

Definition at line 127 of file kerdensom.h.

◆ numNeurons

size_t KerDenSOM::numNeurons

protected

Definition at line 133 of file kerdensom.h.

◆ numVectors

size_t KerDenSOM::numVectors

protected

Definition at line 134 of file kerdensom.h.

◆ reg0

double KerDenSOM::reg0

protected

Definition at line 125 of file kerdensom.h.

◆ reg1

double KerDenSOM::reg1

protected

Definition at line 126 of file kerdensom.h.

◆ sigma

double KerDenSOM::sigma

protected

Definition at line 123 of file kerdensom.h.

◆ somNSteps

size_t KerDenSOM::somNSteps

protected

Definition at line 128 of file kerdensom.h.

◆ tmpD

std::vector<double> KerDenSOM::tmpD

protected

Definition at line 137 of file kerdensom.h.

◆ tmpD1

std::vector<double> KerDenSOM::tmpD1

protected

Definition at line 138 of file kerdensom.h.

◆ tmpDens

std::vector<double> KerDenSOM::tmpDens

protected

Definition at line 139 of file kerdensom.h.

◆ tmpMap

std::vector< std::vector<double> > KerDenSOM::tmpMap

protected

Definition at line 136 of file kerdensom.h.

◆ tmpV

std::vector<double> KerDenSOM::tmpV

protected

Definition at line 140 of file kerdensom.h.

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

xmipp/libraries/classification/kerdensom.h
xmipp/libraries/classification/kerdensom.cpp

Public Member Functions

Protected Member Functions

Protected Attributes

Additional Inherited Members

Detailed Description

Constructor & Destructor Documentation

◆ KerDenSOM()

◆ ~KerDenSOM()

Member Function Documentation

◆ codeDens()

◆ functional()

◆ initU()

◆ mainIterations()

◆ nSteps()

◆ setAnnSteps()

◆ showU()

◆ showV()

◆ showX()

◆ test()

◆ train() [1/2]

◆ train() [2/2]

◆ updateSigmaI()

◆ updateSigmaII()

◆ updateU()

◆ updateU1()

◆ updateV()

◆ updateV1()

Member Data Documentation

◆ annSteps

◆ dim

◆ epsilon

◆ numNeurons

◆ numVectors

◆ reg0

◆ reg1

◆ sigma

◆ somNSteps

◆ tmpD

◆ tmpD1

◆ tmpDens

◆ tmpMap

◆ tmpV