def test_fast_cvm(n_samples=1000):
random = RandomState()
data1 = random.uniform(size=n_samples)
weights1 = random.uniform(size=n_samples)
mask = random.uniform(size=n_samples) > 0.5
data2 = data1[mask]
weights2 = weights1[mask]
a = cvm_2samp(data1, data2, weights1, weights2)
prepared_data1, prepared_weights1, F1 = prepare_distribution(data1, weights1)
b = _cvm_2samp_fast(prepared_data1, data2, prepared_weights1, weights2, cdf1=F1)
assert numpy.allclose(a, b)
def test_cvm2samp(n_samples1=100, n_samples2=100):
data1 = numpy.random.normal(size=n_samples1)
weights1 = numpy.random.random(size=n_samples1)
data2 = numpy.random.normal(size=n_samples2)
weights2 = numpy.random.random(size=n_samples2)
CVM = cvm_2samp(data1, data2, weights1=weights1, weights2=weights2)
# alternative way to check
labels = [0] * len(data1) + [1] * len(data2)
data = numpy.concatenate([data1, data2])
weights = numpy.concatenate([weights1, weights2])
from sklearn.metrics import roc_curve
fpr, tpr, _ = roc_curve(labels, data, sample_weight=weights)
# data1 corresponds to
weights1 = numpy.diff(numpy.insert(fpr, 0, [0]))
CVM2 = numpy.sum(weights1 * (symmetrize(fpr) - symmetrize(tpr)) ** 2)
print(CVM, CVM2)
assert numpy.allclose(CVM, CVM2), 'different values of CVM'
def test_cvm2samp(n_samples1=100, n_samples2=100):
data1 = numpy.random.normal(size=n_samples1)
weights1 = numpy.random.random(size=n_samples1)
data2 = numpy.random.normal(size=n_samples2)
weights2 = numpy.random.random(size=n_samples2)
CVM = cvm_2samp(data1, data2, weights1=weights1, weights2=weights2)
# alternative way to check
labels = [0] * len(data1) + [1] * len(data2)
data = numpy.concatenate([data1, data2])
weights = numpy.concatenate([weights1, weights2])
from sklearn.metrics import roc_curve
fpr, tpr, _ = roc_curve(labels, data, sample_weight=weights)
# data1 corresponds to
weights1 = numpy.diff(numpy.insert(fpr, 0, [0]))
CVM2 = numpy.sum(weights1 * (symmetrize(fpr) - symmetrize(tpr))**2)
print(CVM, CVM2)
assert numpy.allclose(CVM, CVM2), 'different values of CVM'