def check_adaboost_predict(base_estimator, algorithm, expected_score):
train_fs, test_fs = make_sparse_data()
# train an AdaBoostClassifier on the training data and evalute on the
# testing data
learner = Learner('AdaBoostClassifier', model_kwargs={'base_estimator': base_estimator,
'algorithm': algorithm})
learner.train(train_fs, grid_search=False)
test_score = learner.evaluate(test_fs)[1]
assert_almost_equal(test_score, expected_score)
def check_sparse_predict(learner_name, expected_score, use_feature_hashing=False):
train_fs, test_fs = make_sparse_data(
use_feature_hashing=use_feature_hashing)
# train the given classifier on the training
# data and evalute on the testing data
learner = Learner(learner_name)
learner.train(train_fs, grid_search=False)
test_score = learner.evaluate(test_fs)[1]
assert_almost_equal(test_score, expected_score)
def check_sparse_predict_sampler(use_feature_hashing=False):
train_fs, test_fs = make_sparse_data(
use_feature_hashing=use_feature_hashing)
if use_feature_hashing:
sampler = 'RBFSampler'
sampler_parameters = {"gamma": 1.0, "n_components": 50}
else:
sampler = 'Nystroem'
sampler_parameters = {"gamma": 1.0, "n_components": 50,
"kernel": 'rbf'}
learner = Learner('LogisticRegression',
sampler=sampler,
sampler_kwargs=sampler_parameters)
learner.train(train_fs, grid_search=False)
test_score = learner.evaluate(test_fs)[1]
expected_score = 0.48 if use_feature_hashing else 0.45
assert_almost_equal(test_score, expected_score)
def check_sparse_predict_sampler(use_feature_hashing=False):
train_fs, test_fs = make_sparse_data(
use_feature_hashing=use_feature_hashing)
if use_feature_hashing:
sampler = 'RBFSampler'
sampler_parameters = {"gamma": 1.0, "n_components": 50}
else:
sampler = 'Nystroem'
sampler_parameters = {
"gamma": 1.0,
"n_components": 50,
"kernel": 'rbf'
}
learner = Learner('LogisticRegression',
sampler=sampler,
sampler_kwargs=sampler_parameters)
learner.train(train_fs, grid_search=False)
test_score = learner.evaluate(test_fs)[1]
expected_score = 0.48 if use_feature_hashing else 0.45
assert_almost_equal(test_score, expected_score)
