Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
Introduced classifier based on conjugate index
- Loading branch information
Showing
8 changed files
with
444 additions
and
1 deletion.
There are no files selected for viewing
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
42 changes: 42 additions & 0 deletions
42
examples/undocumented/libshogun/classifier_conjugateindex.cpp
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,42 @@ | ||
#include <shogun/features/Labels.h> | ||
#include <shogun/features/SimpleFeatures.h> | ||
#include <shogun/classifier/ConjugateIndex.h> | ||
#include <shogun/base/init.h> | ||
#include <shogun/lib/common.h> | ||
#include <shogun/io/SGIO.h> | ||
|
||
using namespace shogun; | ||
|
||
int main(int argc, char** argv) | ||
{ | ||
init_shogun_with_defaults(); | ||
|
||
// create some data | ||
float64_t* matrix = SG_MALLOC(float64_t, 6); | ||
for (int32_t i=0; i<6; i++) | ||
matrix[i]=i; | ||
|
||
// create three 2-dimensional vectors | ||
// shogun will now own the matrix created | ||
CSimpleFeatures<float64_t>* features= new CSimpleFeatures<float64_t>(); | ||
features->set_feature_matrix(matrix, 2, 3); | ||
|
||
// create three labels | ||
CLabels* labels=new CLabels(3); | ||
labels->set_label(0, 0); | ||
labels->set_label(1, +1); | ||
labels->set_label(2, 0); | ||
|
||
CConjugateIndex* ci = new CConjugateIndex(features,labels); | ||
ci->train(); | ||
|
||
// classify on training examples | ||
for (int32_t i=0; i<3; i++) | ||
SG_SPRINT("output[%d]=%f\n", i, ci->apply(i)); | ||
|
||
// free up memory | ||
SG_UNREF(ci); | ||
|
||
exit_shogun(); | ||
return 0; | ||
} |
27 changes: 27 additions & 0 deletions
27
examples/undocumented/python_modular/classifier_conjugateindex_modular.py
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,27 @@ | ||
from tools.load import LoadMatrix | ||
lm=LoadMatrix() | ||
|
||
traindat = lm.load_numbers('../data/fm_train_real.dat') | ||
testdat = lm.load_numbers('../data/fm_test_real.dat') | ||
label_traindat = lm.load_labels('../data/label_train_multiclass.dat') | ||
|
||
parameter_list = [[traindat,testdat,label_traindat],[traindat,testdat,label_traindat]] | ||
|
||
def classifier_conjugateindex_modular (fm_train_real=traindat,fm_test_real=testdat,label_train_multiclass=label_traindat): | ||
from shogun.Features import RealFeatures, Labels | ||
from shogun.Classifier import ConjugateIndex | ||
|
||
feats_train = RealFeatures(fm_train_real) | ||
feats_test = RealFeatures(fm_test_real) | ||
|
||
labels = Labels(label_train_multiclass) | ||
|
||
ci = ConjugateIndex(feats_train, labels) | ||
ci.train() | ||
|
||
res = ci.apply(feats_test).get_labels() | ||
return ci, res | ||
|
||
if __name__=='__main__': | ||
print 'ConjugateIndex' | ||
classifier_conjugateindex_modular(*parameter_list[0]) |
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Original file line number | Diff line number | Diff line change |
---|---|---|
@@ -0,0 +1,242 @@ | ||
/* | ||
* This program is free software; you can redistribute it and/or modify | ||
* it under the terms of the GNU General Public License as published by | ||
* the Free Software Foundation; either version 3 of the License, or | ||
* (at your option) any later version. | ||
* | ||
* Written (W) 2011 Sergey Lisitsyn | ||
* Copyright (C) 2011 Sergey Lisitsyn | ||
*/ | ||
|
||
#include <shogun/classifier/ConjugateIndex.h> | ||
#include <shogun/machine/Machine.h> | ||
#include <shogun/features/Features.h> | ||
#include <shogun/features/Labels.h> | ||
#include <shogun/mathematics/lapack.h> | ||
#include <shogun/mathematics/Math.h> | ||
#include <shogun/lib/Signal.h> | ||
|
||
using namespace shogun; | ||
|
||
CConjugateIndex::CConjugateIndex() : CMachine() | ||
{ | ||
m_classes = NULL; | ||
m_features = NULL; | ||
}; | ||
|
||
CConjugateIndex::CConjugateIndex(CFeatures* train_features, CLabels* train_labels) : CMachine() | ||
{ | ||
m_features = NULL; | ||
set_features(train_features); | ||
set_labels(train_labels); | ||
m_classes = NULL; | ||
}; | ||
|
||
CConjugateIndex::~CConjugateIndex() | ||
{ | ||
clean_classes(); | ||
m_feature_vector.destroy_vector(); | ||
SG_UNREF(m_features); | ||
}; | ||
|
||
void CConjugateIndex::set_features(CFeatures* features) | ||
{ | ||
ASSERT(features->get_feature_class()==C_SIMPLE); | ||
SG_REF(features); | ||
SG_UNREF(m_features); | ||
m_features = (CSimpleFeatures<float64_t>*)features; | ||
} | ||
|
||
CSimpleFeatures<float64_t>* CConjugateIndex::get_features() | ||
{ | ||
SG_REF(m_features); | ||
return m_features; | ||
} | ||
|
||
void CConjugateIndex::clean_classes() | ||
{ | ||
if (m_classes) | ||
{ | ||
for (int32_t i=0; i<m_num_classes; i++) | ||
m_classes[i].destroy_matrix(); | ||
|
||
delete[] m_classes; | ||
} | ||
} | ||
|
||
bool CConjugateIndex::train(CFeatures* train_features) | ||
{ | ||
if (train_features) | ||
set_features(train_features); | ||
|
||
m_num_classes = labels->get_num_classes(); | ||
ASSERT(m_num_classes>=2); | ||
clean_classes(); | ||
|
||
int32_t num_vectors; | ||
int32_t num_features; | ||
float64_t* feature_matrix = m_features->get_feature_matrix(num_features,num_vectors); | ||
|
||
m_classes = new SGMatrix<float64_t>[m_num_classes]; | ||
for (int32_t i=0; i<m_num_classes; i++) | ||
m_classes[i] = SGMatrix<float64_t>(num_features,num_features); | ||
|
||
m_feature_vector = SGVector<float64_t>(num_features); | ||
|
||
//float64_t* evals = SG_MALLOC(float64_t, num_features); | ||
//float64_t* evecs = SG_MALLOC(float64_t, num_features*num_features); | ||
SGMatrix<float64_t> matrix(CMath::max(num_features,num_vectors),CMath::max(num_features,num_vectors)); | ||
SGMatrix<float64_t> class_feature_matrix(num_features,CMath::max(num_features,num_vectors)); | ||
SGMatrix<float64_t> helper_matrix(CMath::max(num_features,num_vectors),num_features); | ||
|
||
SG_PROGRESS(0,0,m_num_classes-1); | ||
for (int32_t label=0; label<m_num_classes; label++) | ||
{ | ||
/* | ||
int32_t count = 0; | ||
for (int32_t i=0; i<num_vectors; i++) | ||
{ | ||
if ((int32_t)labels->get_label(i) == label) | ||
{ | ||
for (int32_t j=0; j<num_features; j++) | ||
{ | ||
for (int32_t k=0; k<num_features; k++) | ||
{ | ||
matrix[j*num_features+k] += | ||
feature_matrix[i*num_features+j]* | ||
feature_matrix[i*num_features+k]; | ||
} | ||
} | ||
count++; | ||
} | ||
} | ||
ASSERT(num_features>count); | ||
int32_t info = 0; | ||
wrap_dsyevr('V','U',num_features,matrix.matrix,num_features,1,num_features-count+1,evals,evecs,&info); | ||
cblas_dgemm(CblasColMajor,CblasNoTrans,CblasTrans, | ||
num_features,num_features,num_features-count-1, | ||
1.0,evecs,num_features, | ||
evecs,num_features, | ||
0.0,m_classes[label].matrix,num_features); | ||
ASSERT(!info); | ||
*/ | ||
int32_t count = 0; | ||
for (int32_t i=0; i<num_vectors; i++) | ||
{ | ||
if ((int32_t)labels->get_label(i) == label) | ||
count++; | ||
} | ||
|
||
count = 0; | ||
for (int32_t i=0; i<num_vectors; i++) | ||
{ | ||
if (labels->get_label(i) == label) | ||
{ | ||
memcpy(class_feature_matrix.matrix+count*num_features, | ||
feature_matrix+i*num_features, | ||
sizeof(float64_t)*num_features); | ||
count++; | ||
} | ||
} | ||
|
||
cblas_dgemm(CblasColMajor,CblasTrans,CblasNoTrans, | ||
count,count,num_features, | ||
1.0,class_feature_matrix.matrix,num_features, | ||
class_feature_matrix.matrix,num_features, | ||
0.0,matrix.matrix,count); | ||
|
||
CMath::inverse(SGMatrix<float64_t>(matrix.matrix,count,count)); | ||
|
||
cblas_dgemm(CblasColMajor,CblasNoTrans,CblasTrans, | ||
count,num_features,count, | ||
1.0,matrix.matrix,count, | ||
class_feature_matrix.matrix,num_features, | ||
0.0,helper_matrix.matrix,count); | ||
|
||
cblas_dgemm(CblasColMajor,CblasNoTrans,CblasNoTrans, | ||
num_features,num_features,count, | ||
1.0,class_feature_matrix.matrix,num_features, | ||
helper_matrix.matrix,count, | ||
0.0,m_classes[label].matrix,num_features); | ||
|
||
SG_PROGRESS(label+1,0,m_num_classes); | ||
} | ||
helper_matrix.destroy_matrix(); | ||
class_feature_matrix.destroy_matrix(); | ||
matrix.destroy_matrix(); | ||
SG_DONE(); | ||
|
||
return true; | ||
}; | ||
|
||
CLabels* CConjugateIndex::apply(CFeatures* test_features) | ||
{ | ||
set_features(test_features); | ||
|
||
CLabels* predicted_labels = apply(); | ||
|
||
return predicted_labels; | ||
}; | ||
|
||
CLabels* CConjugateIndex::apply() | ||
{ | ||
ASSERT(m_classes); | ||
ASSERT(m_num_classes>1); | ||
ASSERT(m_features->get_num_features()==m_feature_vector.vlen); | ||
|
||
int32_t num_vectors = m_features->get_num_vectors(); | ||
|
||
CLabels* predicted_labels = new CLabels(num_vectors); | ||
|
||
for (int32_t i=0; i<num_vectors;i++) | ||
{ | ||
SG_PROGRESS(i,0,num_vectors-1); | ||
predicted_labels->set_label(i,apply(i)); | ||
} | ||
SG_DONE(); | ||
|
||
return predicted_labels; | ||
}; | ||
|
||
float64_t CConjugateIndex::conjugate_index(SGVector<float64_t> feature_vector, int32_t label) | ||
{ | ||
int32_t num_features = feature_vector.vlen; | ||
float64_t norm = cblas_ddot(num_features,feature_vector.vector,1, | ||
feature_vector.vector,1); | ||
|
||
cblas_dgemv(CblasColMajor,CblasNoTrans, | ||
num_features,num_features, | ||
1.0,m_classes[label].matrix,num_features, | ||
feature_vector.vector,1, | ||
0.0,m_feature_vector.vector,1); | ||
|
||
float64_t product = cblas_ddot(num_features,feature_vector.vector,1, | ||
m_feature_vector.vector,1); | ||
return product/norm; | ||
}; | ||
|
||
float64_t CConjugateIndex::apply(int32_t index) | ||
{ | ||
int32_t predicted_label = 0; | ||
float64_t max_conjugate_index = 0.0; | ||
float64_t current_conjugate_index; | ||
|
||
SGVector<float64_t> feature_vector = m_features->get_feature_vector(index); | ||
for (int32_t i=0; i<m_num_classes; i++) | ||
{ | ||
current_conjugate_index = conjugate_index(feature_vector,i); | ||
|
||
if (current_conjugate_index > max_conjugate_index) | ||
{ | ||
max_conjugate_index = current_conjugate_index; | ||
predicted_label = i; | ||
} | ||
} | ||
feature_vector.free_vector(); | ||
|
||
return predicted_label; | ||
}; | ||
|
||
|
||
|
Oops, something went wrong.