def test_build_ratings_matrix():
collaborative_filtering = CollaborativeFiltering(7)
Y, users, movies = collaborative_filtering._build_ratings_matrix(
to_numpy(X_train), to_numpy((y_train)))
assert_array_equal(Y, Y_expected)
assert_array_equal(users, users_expected)
assert_array_equal(movies, movies_expected)
def test_gradient():
x = to_numpy([5, -9, 12])
numerical_grad = gradient(x, f)
analytical_grad = grad_f(x)
assert_allclose(numerical_grad, analytical_grad)
def _check_X_y(self, X, y):
X = self._check_X(X)
y = to_numpy(y)
if y.ndim != 1:
raise ValueError(
f"Expected {(X.shape[0],)} shape of ratings, but {y.shape} received"
)
return X, y
def _check_X(self, X):
X = to_numpy(X)
if X.ndim != 2:
raise ValueError(
"User-item pairs (X) should be a two-dimensional array")
if X.shape[1] != 2:
raise ValueError(
f"Expected 2 columns in user-item pairs (X), but {X.shape[1]} received"
)
return X
def test_fit_predict():
model = KMeans(3)
model.fit(X)
predicted_clusters_labels = model.predict(X)
X_np = to_numpy(X)
actual_clusters = [
X_np[actual_clusters_labels == label].tolist()
for label in np.unique(actual_clusters_labels)
]
for label in np.unique(predicted_clusters_labels):
predicted_cluster = X_np[predicted_clusters_labels == label].tolist()
assert predicted_cluster in actual_clusters
def test_cost_gradient(features_count, regularization_param, Y):
Y = to_numpy(Y)
users_count = Y.shape[1]
items_count = Y.shape[0]
params = unroll(
glorot_init(
((users_count, features_count), (items_count, features_count))))
collaborative_filtering = CollaborativeFiltering(features_count,
regularization_param)
collaborative_filtering._users_count = users_count
collaborative_filtering._items_count = items_count
analytical_gradient = collaborative_filtering._cost_gradient(params, Y)
numerical_gradient = gradient(params, collaborative_filtering._cost, (Y, ))
assert_allclose(analytical_gradient,
numerical_gradient,
rtol=1E-4,
atol=1E-4)
"""Tests for the KMeans class."""
import pytest
import numpy as np
from mymllib.clustering import KMeans
from mymllib.preprocessing import to_numpy
X = [[5, 0], [4, 1], [6, 2], [0, 6], [1, 5], [2, 7], [6, 6], [7, 5], [4, 7]]
actual_clusters_labels = [0, 0, 0, 1, 1, 1, 2, 2, 2]
@pytest.mark.parametrize("X", [to_numpy(X)])
@pytest.mark.parametrize("clusters_count", [len(X), len(X) + 1])
def test_random_init__not_enough_samples(X, clusters_count):
model = KMeans(clusters_count)
with pytest.raises(ValueError):
model._random_init(X)
@pytest.mark.parametrize("X", [to_numpy(X)])
@pytest.mark.parametrize("clusters_count", [1, 2, 3, len(X) - 1])
def test_random_init(X, clusters_count):
model = KMeans(clusters_count)
cluster_centroids = model._random_init(X)
# Check that correct number of centroids was returned
assert cluster_centroids.shape[0] == clusters_count
# Check that all centroids are unique
assert np.unique(cluster_centroids,
def test_minimize(optimizer):
x0 = to_numpy([-7, 15, 4])
x = optimizer.minimize(f, grad_f, x0)
assert_allclose(x, min_f, atol=1E-5)
def grad_f(x):
return to_numpy([4*x[0], 2*x[1], 260*x[2]])
"""Tests for optimizers (subclasses of the BaseOptimizer class)."""
import pytest
from numpy.testing import assert_allclose
from mymllib.optimization import GradientDescent, SciPyOptimizer
from mymllib.preprocessing import to_numpy
def f(x):
return 2*x[0]**2 + 0.5*x[1]**4 + 130*x[2]**2
def grad_f(x):
return to_numpy([4*x[0], 2*x[1], 260*x[2]])
min_f = to_numpy([0, 0, 0]) # Minimum of f()
@pytest.mark.parametrize("optimizer", [GradientDescent(max_iterations=10000), SciPyOptimizer("L-BFGS-B")])
def test_minimize(optimizer):
x0 = to_numpy([-7, 15, 4])
x = optimizer.minimize(f, grad_f, x0)
assert_allclose(x, min_f, atol=1E-5)
def test_to_numpy__one_arg_passed():
A_numpy = to_numpy(A)
assert isinstance(A_numpy, ndarray)
def test_one_hot(y, expected_labels, expected_y_one_hot):
labels, y_one_hot = one_hot(to_numpy(y))
assert_array_equal(labels, expected_labels)
assert_array_equal(y_one_hot, expected_y_one_hot)
def test_to_numpy__two_args_passed():
A_numpy, B_numpy = to_numpy(A, [1, 2, 3])
assert isinstance(A_numpy, ndarray)
assert isinstance(B_numpy, ndarray)