Skip to content

Commit

Permalink
Add 7 Tests.
Browse files Browse the repository at this point in the history
  • Loading branch information
@pl8787
pl8787 committed Mar 6, 2014
1 parent 51fdd04 commit a51badb
Showing 1 changed file with 288 additions and 60 deletions.
348 changes: 288 additions & 60 deletions tests/unit/classifier/svm/LibLinear_unittest.cc
View file
Expand Up @@ -8,8 +8,9 @@
using namespace shogun;

#ifdef HAVE_LAPACK
TEST(LibLinearTest,train)
TEST(LibLinearTest,train_L2R_LR)
{
LIBLINEAR_SOLVER_TYPE liblinear_solver_type = L2R_LR;
index_t num_samples = 50;
CMath::init_random(5);
SGMatrix<float64_t> data =
Expand All @@ -30,40 +31,104 @@ TEST(LibLinearTest,train)

CDenseFeatures<float64_t>* train_feats = (CDenseFeatures<float64_t>*)features.copy_subset(train_idx);
CDenseFeatures<float64_t>* test_feats = (CDenseFeatures<float64_t>*)features.copy_subset(test_idx);
CBinaryLabels* ground_truth = new CBinaryLabels(labels);

CLibLinear* ll = new CLibLinear();

CBinaryLabels* pred = NULL;
float64_t liblin_accuracy;
CContingencyTableEvaluation* eval = new CContingencyTableEvaluation();

ll->set_bias_enabled(true);
ll->set_features((CDenseFeatures<float64_t> *)(train_feats->clone()));
ll->set_labels(ground_truth);

ll->set_liblinear_solver_type(liblinear_solver_type);
ll->train();
pred = ll->apply_binary(test_feats);
liblin_accuracy = eval->evaluate(pred, ground_truth);

CDenseFeatures<float64_t>* train_feats_l1 = (CDenseFeatures<float64_t>*)features.copy_subset(train_idx);
CDenseFeatures<float64_t>* test_feats_l1 = (CDenseFeatures<float64_t>*)features.copy_subset(test_idx);
EXPECT_NEAR(liblin_accuracy, 1.0, 1e-6);

SGMatrix<float64_t> train_matrix = train_feats_l1->get_feature_matrix();
SGMatrix<float64_t> test_matrix = test_feats_l1->get_feature_matrix();
SG_UNREF(train_feats);
SG_UNREF(test_feats);
SG_UNREF(pred);
}

TEST(LibLinearTest,train_L2R_L2LOSS_SVC_DUAL)
{
LIBLINEAR_SOLVER_TYPE liblinear_solver_type = L2R_L2LOSS_SVC_DUAL;
index_t num_samples = 50;
CMath::init_random(5);
SGMatrix<float64_t> data =
CDataGenerator::generate_gaussians(num_samples, 2, 2);
CDenseFeatures<float64_t> features(data);

SGMatrix<float64_t>::transpose_matrix(train_matrix.matrix, train_matrix.num_rows, train_matrix.num_cols);
SGMatrix<float64_t>::transpose_matrix(test_matrix.matrix, test_matrix.num_rows, test_matrix.num_cols);
SGVector<index_t> train_idx(num_samples), test_idx(num_samples);
SGVector<float64_t> labels(num_samples);
for (index_t i = 0, j = 0; i < data.num_cols; ++i)
{
if (i % 2 == 0)
train_idx[j] = i;
else
test_idx[j++] = i;

train_feats_l1 = new CDenseFeatures<float64_t>(train_matrix);
test_feats_l1 = new CDenseFeatures<float64_t>(test_matrix);
labels[i/2] = (i < data.num_cols/2) ? 1.0 : -1.0;
}

CDenseFeatures<float64_t>* train_feats = (CDenseFeatures<float64_t>*)features.copy_subset(train_idx);
CDenseFeatures<float64_t>* test_feats = (CDenseFeatures<float64_t>*)features.copy_subset(test_idx);
CBinaryLabels* ground_truth = new CBinaryLabels(labels);

LIBLINEAR_SOLVER_TYPE ll_list[7] = {
/// L2 regularized linear logistic regression
L2R_LR,
/// L2 regularized SVM with L2-loss using dual coordinate descent
L2R_L2LOSS_SVC_DUAL,
/// L2 regularized SVM with L2-loss using newton in the primal
L2R_L2LOSS_SVC,
/// L2 regularized linear SVM with L1-loss using dual coordinate descent
L2R_L1LOSS_SVC_DUAL,
/// L1 regularized SVM with L2-loss using dual coordinate descent
L1R_L2LOSS_SVC,
/// L1 regularized logistic regression
L1R_LR,
/// L2 regularized linear logistic regression via dual
L2R_LR_DUAL
};
CLibLinear* ll = new CLibLinear();

CBinaryLabels* pred = NULL;
float64_t liblin_accuracy;
CContingencyTableEvaluation* eval = new CContingencyTableEvaluation();

ll->set_bias_enabled(true);
ll->set_features((CDenseFeatures<float64_t> *)(train_feats->clone()));
ll->set_labels(ground_truth);

ll->set_liblinear_solver_type(liblinear_solver_type);
ll->train();
pred = ll->apply_binary(test_feats);
liblin_accuracy = eval->evaluate(pred, ground_truth);

EXPECT_NEAR(liblin_accuracy, 1.0, 1e-6);

SG_UNREF(train_feats);
SG_UNREF(test_feats);
SG_UNREF(pred);
}

TEST(LibLinearTest,train_L2R_L2LOSS_SVC)
{
LIBLINEAR_SOLVER_TYPE liblinear_solver_type = L2R_L2LOSS_SVC;
index_t num_samples = 50;
CMath::init_random(5);
SGMatrix<float64_t> data =
CDataGenerator::generate_gaussians(num_samples, 2, 2);
CDenseFeatures<float64_t> features(data);

SGVector<index_t> train_idx(num_samples), test_idx(num_samples);
SGVector<float64_t> labels(num_samples);
for (index_t i = 0, j = 0; i < data.num_cols; ++i)
{
if (i % 2 == 0)
train_idx[j] = i;
else
test_idx[j++] = i;

labels[i/2] = (i < data.num_cols/2) ? 1.0 : -1.0;
}

CDenseFeatures<float64_t>* train_feats = (CDenseFeatures<float64_t>*)features.copy_subset(train_idx);
CDenseFeatures<float64_t>* test_feats = (CDenseFeatures<float64_t>*)features.copy_subset(test_idx);
CBinaryLabels* ground_truth = new CBinaryLabels(labels);

CLibLinear* ll = new CLibLinear();
CLibLinear* ll_l1 = new CLibLinear();

CBinaryLabels* pred = NULL;
float64_t liblin_accuracy;
CContingencyTableEvaluation* eval = new CContingencyTableEvaluation();
Expand All @@ -72,48 +137,211 @@ TEST(LibLinearTest,train)
ll->set_features((CDenseFeatures<float64_t> *)(train_feats->clone()));
ll->set_labels(ground_truth);

ll_l1->set_bias_enabled(true);
ll_l1->set_features((CDenseFeatures<float64_t> *)(train_feats_l1->clone()));
ll_l1->set_labels(ground_truth);
ll->set_liblinear_solver_type(liblinear_solver_type);
ll->train();
pred = ll->apply_binary(test_feats);
liblin_accuracy = eval->evaluate(pred, ground_truth);

EXPECT_NEAR(liblin_accuracy, 1.0, 1e-6);

SG_UNREF(train_feats);
SG_UNREF(test_feats);
SG_UNREF(pred);
}

TEST(LibLinearTest,train_L2R_L1LOSS_SVC_DUAL)
{
LIBLINEAR_SOLVER_TYPE liblinear_solver_type = L2R_L1LOSS_SVC_DUAL;
index_t num_samples = 50;
CMath::init_random(5);
SGMatrix<float64_t> data =
CDataGenerator::generate_gaussians(num_samples, 2, 2);
CDenseFeatures<float64_t> features(data);

for(index_t i = 0; i < 7; i++)
SGVector<index_t> train_idx(num_samples), test_idx(num_samples);
SGVector<float64_t> labels(num_samples);
for (index_t i = 0, j = 0; i < data.num_cols; ++i)
{
LIBLINEAR_SOLVER_TYPE liblinear_solver_type = ll_list[i];
SG_SPRINT("Begin LibLinear %d \n", liblinear_solver_type);
if(liblinear_solver_type == L1R_L2LOSS_SVC ||
(liblinear_solver_type == L1R_LR) )
if (i % 2 == 0)
train_idx[j] = i;
else
test_idx[j++] = i;

labels[i/2] = (i < data.num_cols/2) ? 1.0 : -1.0;
}

CDenseFeatures<float64_t>* train_feats = (CDenseFeatures<float64_t>*)features.copy_subset(train_idx);
CDenseFeatures<float64_t>* test_feats = (CDenseFeatures<float64_t>*)features.copy_subset(test_idx);
CBinaryLabels* ground_truth = new CBinaryLabels(labels);

CLibLinear* ll = new CLibLinear();

CBinaryLabels* pred = NULL;
float64_t liblin_accuracy;
CContingencyTableEvaluation* eval = new CContingencyTableEvaluation();

ll->set_bias_enabled(true);
ll->set_features((CDenseFeatures<float64_t> *)(train_feats->clone()));
ll->set_labels(ground_truth);

ll->set_liblinear_solver_type(liblinear_solver_type);
ll->train();
pred = ll->apply_binary(test_feats);
liblin_accuracy = eval->evaluate(pred, ground_truth);

EXPECT_NEAR(liblin_accuracy, 1.0, 1e-6);

SG_UNREF(train_feats);
SG_UNREF(test_feats);
SG_UNREF(pred);
}

TEST(LibLinearTest,train_L1R_L2LOSS_SVC)
{
LIBLINEAR_SOLVER_TYPE liblinear_solver_type = L1R_L2LOSS_SVC;
index_t num_samples = 50;
CMath::init_random(5);
SGMatrix<float64_t> data =
CDataGenerator::generate_gaussians(num_samples, 2, 2);
CDenseFeatures<float64_t> features(data);

SGVector<index_t> train_idx(num_samples), test_idx(num_samples);
SGVector<float64_t> labels(num_samples);
for (index_t i = 0, j = 0; i < data.num_cols; ++i)
{
SG_SPRINT("num_fea: %d\tnum_vec: %d\n",
ll_l1->get_features()->get_dim_feature_space(),
ll_l1->get_features()->get_num_vectors());
ll_l1->set_liblinear_solver_type(ll_list[i]);
SG_SPRINT("Begin of L1 Training %d.\n", liblinear_solver_type);
ll_l1->train();
SG_SPRINT("End of L1 Training %d.\n", liblinear_solver_type);
pred = ll_l1->apply_binary(test_feats);
liblin_accuracy = eval->evaluate(pred, ground_truth);
SG_SPRINT("LibLinear(%d) accuracy: %f\n", liblinear_solver_type, liblin_accuracy);
EXPECT_NEAR(liblin_accuracy, 1.0, 1e-6);
SG_SPRINT("num_fea: %d\tnum_vec: %d\n",
ll_l1->get_features()->get_dim_feature_space(),
ll_l1->get_features()->get_num_vectors());
ll_l1->set_features((CDenseFeatures<float64_t> *)(train_feats_l1->clone()));
if (i % 2 == 0)
train_idx[j] = i;
else
test_idx[j++] = i;

labels[i/2] = (i < data.num_cols/2) ? 1.0 : -1.0;
}
else

CDenseFeatures<float64_t>* train_feats = (CDenseFeatures<float64_t>*)features.copy_subset(train_idx);
CDenseFeatures<float64_t>* test_feats = (CDenseFeatures<float64_t>*)features.copy_subset(test_idx);

SGMatrix<float64_t> train_matrix = train_feats->get_feature_matrix();
SGMatrix<float64_t>::transpose_matrix(train_matrix.matrix, train_matrix.num_rows, train_matrix.num_cols);
train_feats = new CDenseFeatures<float64_t>(train_matrix);

CBinaryLabels* ground_truth = new CBinaryLabels(labels);

CLibLinear* ll = new CLibLinear();

CBinaryLabels* pred = NULL;
float64_t liblin_accuracy;
CContingencyTableEvaluation* eval = new CContingencyTableEvaluation();

ll->set_bias_enabled(true);
ll->set_features((CDenseFeatures<float64_t> *)(train_feats->clone()));
ll->set_labels(ground_truth);

ll->set_liblinear_solver_type(liblinear_solver_type);
ll->train();
pred = ll->apply_binary(test_feats);
liblin_accuracy = eval->evaluate(pred, ground_truth);

EXPECT_NEAR(liblin_accuracy, 1.0, 1e-6);

SG_UNREF(train_feats);
SG_UNREF(test_feats);
SG_UNREF(pred);
}

TEST(LibLinearTest,train_L1R_LR)
{
LIBLINEAR_SOLVER_TYPE liblinear_solver_type = L1R_LR;
index_t num_samples = 50;
CMath::init_random(5);
SGMatrix<float64_t> data =
CDataGenerator::generate_gaussians(num_samples, 2, 2);
CDenseFeatures<float64_t> features(data);

SGVector<index_t> train_idx(num_samples), test_idx(num_samples);
SGVector<float64_t> labels(num_samples);
for (index_t i = 0, j = 0; i < data.num_cols; ++i)
{
ll->set_liblinear_solver_type(ll_list[i]);
SG_SPRINT("Begin of Training %d.\n", liblinear_solver_type);
ll->train();
SG_SPRINT("End of Training %d.\n", liblinear_solver_type);
pred = ll->apply_binary(test_feats);
liblin_accuracy = eval->evaluate(pred, ground_truth);
SG_SPRINT("LibLinear(%d) accuracy: %f\n", liblinear_solver_type, liblin_accuracy);
EXPECT_NEAR(liblin_accuracy, 1.0, 1e-6);
ll->set_features((CDenseFeatures<float64_t> *)(train_feats->clone()));
if (i % 2 == 0)
train_idx[j] = i;
else
test_idx[j++] = i;

labels[i/2] = (i < data.num_cols/2) ? 1.0 : -1.0;
}
SG_SPRINT("\n");

CDenseFeatures<float64_t>* train_feats = (CDenseFeatures<float64_t>*)features.copy_subset(train_idx);
CDenseFeatures<float64_t>* test_feats = (CDenseFeatures<float64_t>*)features.copy_subset(test_idx);

SGMatrix<float64_t> train_matrix = train_feats->get_feature_matrix();
SGMatrix<float64_t>::transpose_matrix(train_matrix.matrix, train_matrix.num_rows, train_matrix.num_cols);
train_feats = new CDenseFeatures<float64_t>(train_matrix);

CBinaryLabels* ground_truth = new CBinaryLabels(labels);

CLibLinear* ll = new CLibLinear();

CBinaryLabels* pred = NULL;
float64_t liblin_accuracy;
CContingencyTableEvaluation* eval = new CContingencyTableEvaluation();

ll->set_bias_enabled(true);
ll->set_features((CDenseFeatures<float64_t> *)(train_feats->clone()));
ll->set_labels(ground_truth);

ll->set_liblinear_solver_type(liblinear_solver_type);
ll->train();
pred = ll->apply_binary(test_feats);
liblin_accuracy = eval->evaluate(pred, ground_truth);

EXPECT_NEAR(liblin_accuracy, 1.0, 1e-6);

SG_UNREF(train_feats);
SG_UNREF(test_feats);
SG_UNREF(pred);
}

TEST(LibLinearTest,train_L2R_LR_DUAL)
{
LIBLINEAR_SOLVER_TYPE liblinear_solver_type = L2R_LR_DUAL;
index_t num_samples = 50;
CMath::init_random(5);
SGMatrix<float64_t> data =
CDataGenerator::generate_gaussians(num_samples, 2, 2);
CDenseFeatures<float64_t> features(data);

SGVector<index_t> train_idx(num_samples), test_idx(num_samples);
SGVector<float64_t> labels(num_samples);
for (index_t i = 0, j = 0; i < data.num_cols; ++i)
{
if (i % 2 == 0)
train_idx[j] = i;
else
test_idx[j++] = i;

labels[i/2] = (i < data.num_cols/2) ? 1.0 : -1.0;
}

CDenseFeatures<float64_t>* train_feats = (CDenseFeatures<float64_t>*)features.copy_subset(train_idx);
CDenseFeatures<float64_t>* test_feats = (CDenseFeatures<float64_t>*)features.copy_subset(test_idx);
CBinaryLabels* ground_truth = new CBinaryLabels(labels);

CLibLinear* ll = new CLibLinear();

CBinaryLabels* pred = NULL;
float64_t liblin_accuracy;
CContingencyTableEvaluation* eval = new CContingencyTableEvaluation();

ll->set_bias_enabled(true);
ll->set_features((CDenseFeatures<float64_t> *)(train_feats->clone()));
ll->set_labels(ground_truth);

ll->set_liblinear_solver_type(liblinear_solver_type);
ll->train();
pred = ll->apply_binary(test_feats);
liblin_accuracy = eval->evaluate(pred, ground_truth);

EXPECT_NEAR(liblin_accuracy, 1.0, 1e-6);

SG_UNREF(train_feats);
SG_UNREF(test_feats);
SG_UNREF(pred);
Expand Down

0 comments on commit a51badb

Please sign in to comment.