def test_pybrain_reproducibility():
# This test fails. Because PyBrain can't reproduce training.
X, y, _ = generate_classification_data()
clf1 = PyBrainClassifier(layers=[4], epochs=2).fit(X, y)
clf2 = PyBrainClassifier(layers=[4], epochs=2).fit(X, y)
print(clf1.predict_proba(X) - clf2.predict_proba(X))
assert numpy.allclose(clf1.predict_proba(X), clf2.predict_proba(X)), 'different predicitons'
check_classification_reproducibility(clf1, X, y)
def test_theanets_reproducibility():
clf = TheanetsClassifier(trainers=[{
'algo': 'nag',
'min_improvement': 0.1,
'max_updates': 10
}])
X, y, _ = generate_classification_data()
check_classification_reproducibility(clf, X, y)
def test_theanets_reproducibility():
clf = TheanetsClassifier(trainers=[{
'algo': 'nag',
'min_improvement': 0.1
}])
X, y, _ = generate_classification_data()
import numpy
numpy.random.seed(43)
check_classification_reproducibility(clf, X, y)
def test_pybrain_reproducibility():
# This test fails. Because PyBrain can't reproduce training.
X, y, _ = generate_classification_data()
clf1 = PyBrainClassifier(layers=[4], epochs=2).fit(X, y)
clf2 = PyBrainClassifier(layers=[4], epochs=2).fit(X, y)
print(clf1.predict_proba(X) - clf2.predict_proba(X))
assert numpy.allclose(clf1.predict_proba(X),
clf2.predict_proba(X)), 'different predicitons'
check_classification_reproducibility(clf1, X, y)
def test_mn_reproducibility():
clf = MatrixNetClassifier(iterations=10)
X, y, _ = generate_classification_data()
check_classification_reproducibility(clf, X, y)
def test_neurolab_reproducibility():
clf = NeurolabClassifier(layers=[4, 5], epochs=2, trainf=nl.train.train_gd)
X, y, _ = generate_classification_data()
check_classification_reproducibility(clf, X, y)
def test_neurolab_reproducibility():
clf = NeurolabClassifier(layers=[4, 5], epochs=2, trainf=nl.train.train_gd)
X, y, _ = generate_classification_data()
check_classification_reproducibility(clf, X, y)
def test_theanets_reproducibility():
clf = TheanetsClassifier(trainers=[{'min_improvement': 1}])
X, y, _ = generate_classification_data()
check_classification_reproducibility(clf, X, y)
def test_theanets_reproducibility():
clf = TheanetsClassifier(trainers=[{'min_improvement': 1}])
X, y, _ = generate_classification_data()
check_classification_reproducibility(clf, X, y)
def test_mn_reproducibility():
clf = MatrixNetClassifier(iterations=10)
X, y, _ = generate_classification_data()
check_classification_reproducibility(clf, X, y)
def test_theanets_reproducibility():
clf = TheanetsClassifier(trainers=[{'algo': 'nag', 'min_improvement': 0.1}])
X, y, _ = generate_classification_data()
import numpy
numpy.random.seed(43)
check_classification_reproducibility(clf, X, y)
def test_theanets_reproducibility():
clf = TheanetsClassifier(trainers=[{'algo': 'nag', 'min_improvement': 0.1, 'max_updates': 10}])
X, y, _ = generate_classification_data()
check_classification_reproducibility(clf, X, y)
