def test_clusters(self):
"""Tests the shape of the clusters."""
gmm = GMM(self.num_centers,
initial_clusters=self.initial_means,
batch_size=self.batch_size,
steps=40,
continue_training=True,
random_seed=4,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(x=self.points, steps=0)
clusters = gmm.clusters()
self.assertAllEqual(list(clusters.shape),
[self.num_centers, self.num_dims])
def test_clusters(self):
"""Tests the shape of the clusters."""
gmm = GMM(self.num_centers,
initial_clusters=self.initial_means,
batch_size=self.batch_size,
steps=40,
continue_training=True,
random_seed=4,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(x=self.points, steps=0)
clusters = gmm.clusters()
self.assertAllEqual(list(clusters.shape),
[self.num_centers, self.num_dims])
def _compare_with_sklearn(self, cov_type):
# sklearn version.
iterations = 40
np.random.seed(5)
sklearn_assignments = np.asarray([0, 0, 1, 0, 0, 0, 1, 0, 0, 1])
sklearn_means = np.asarray([[144.83417719, 254.20130341],
[274.38754816, 353.16074346]])
sklearn_covs = np.asarray([[[395.0081194, -4.50389512],
[-4.50389512, 408.27543989]],
[[385.17484203, -31.27834935],
[-31.27834935, 391.74249925]]])
# skflow version.
gmm = GMM(self.num_centers,
initial_clusters=self.initial_means,
covariance_type=cov_type,
batch_size=self.num_points,
steps=iterations,
continue_training=True,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(self.points)
skflow_assignments = gmm.predict(self.points[:10, :]).astype(int)
self.assertAllClose(sklearn_assignments, np.ravel(skflow_assignments))
self.assertAllClose(sklearn_means, gmm.clusters())
if cov_type == 'full':
self.assertAllClose(sklearn_covs, gmm.covariances(), rtol=0.01)
else:
for d in [0, 1]:
self.assertAllClose(np.diag(sklearn_covs[d]),
gmm.covariances()[d, :],
rtol=0.01)
def test_fit(self):
gmm = GMM(self.num_centers,
initial_clusters='random',
batch_size=self.batch_size,
random_seed=4,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(x=self.points, steps=1)
score1 = gmm.score(x=self.points)
gmm = GMM(self.num_centers,
initial_clusters='random',
batch_size=self.batch_size,
random_seed=4,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(x=self.points, steps=10)
score2 = gmm.score(x=self.points)
self.assertGreater(score1, score2)
self.assertNear(self.true_score, score2, self.true_score * 0.15)
def _compare_with_sklearn(self, cov_type):
# sklearn version.
iterations = 40
np.random.seed(5)
sklearn_assignments = np.asarray([0, 0, 1, 0, 0, 0, 1, 0, 0, 1])
sklearn_means = np.asarray([[144.83417719, 254.20130341],
[274.38754816, 353.16074346]])
sklearn_covs = np.asarray([[[395.0081194, -4.50389512],
[-4.50389512, 408.27543989]],
[[385.17484203, -31.27834935],
[-31.27834935, 391.74249925]]])
# skflow version.
gmm = GMM(self.num_centers,
initial_clusters=self.initial_means,
covariance_type=cov_type,
batch_size=self.num_points,
steps=iterations,
continue_training=True,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(self.points)
skflow_assignments = gmm.predict(self.points[:10, :]).astype(int)
self.assertAllClose(sklearn_assignments,
np.ravel(skflow_assignments))
self.assertAllClose(sklearn_means, gmm.clusters())
if cov_type == 'full':
self.assertAllClose(sklearn_covs, gmm.covariances(), rtol=0.01)
else:
for d in [0, 1]:
self.assertAllClose(np.diag(sklearn_covs[d]),
gmm.covariances()[d, :], rtol=0.01)
def test_fit(self):
gmm = GMM(self.num_centers,
initial_clusters='random',
batch_size=self.batch_size,
random_seed=4,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(x=self.points, steps=1)
score1 = gmm.score(x=self.points)
gmm = GMM(self.num_centers,
initial_clusters='random',
batch_size=self.batch_size,
random_seed=4,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(x=self.points, steps=10)
score2 = gmm.score(x=self.points)
self.assertGreater(score1, score2)
self.assertNear(self.true_score, score2, self.true_score * 0.15)
def main():
img = mpimg.imread('corgi.png')[:, :, :3]
img_reshape = np.reshape(img, (-1, 3))
for m in [3, 5, 10]:
gmm = GMM(m, steps=10)
gmm.fit(img_reshape)
print("w:")
print(weights(gmm))
print("mu:")
print(gmm.clusters())
print("sigma:")
print(gmm.covariances())
clusters = gmm.clusters()
img_cluster = clusters[gmm.predict(img_reshape)]
img_cluster = np.reshape(img_cluster,
(img.shape[0], img.shape[1], img.shape[2]))
plt.imshow(img_cluster)
plt.title("m = %d" % m)
plt.show()
def test_infer(self):
gmm = GMM(self.num_centers,
initial_clusters=self.initial_means,
batch_size=self.batch_size,
steps=40,
continue_training=True,
random_seed=4,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(x=self.points, steps=60)
clusters = gmm.clusters()
# Make a small test set
points, true_assignments, true_offsets = (self.make_random_points(
clusters, 40))
assignments = np.ravel(gmm.predict(points))
self.assertAllEqual(true_assignments, assignments)
# Test score
score = gmm.score(points)
self.assertNear(score, np.sum(true_offsets), 4.05)
def test_infer(self):
gmm = GMM(self.num_centers,
initial_clusters=self.initial_means,
batch_size=self.batch_size,
steps=40,
continue_training=True,
random_seed=4,
config=run_config.RunConfig(tf_random_seed=2))
gmm.fit(x=self.points, steps=60)
clusters = gmm.clusters()
# Make a small test set
points, true_assignments, true_offsets = (
self.make_random_points(clusters, 40))
assignments = np.ravel(gmm.predict(points))
self.assertAllEqual(true_assignments, assignments)
# Test score
score = gmm.score(points)
self.assertNear(score, np.sum(true_offsets), 4.05)
