def testMultipleRuns(self):
"""testing 50 runs with 10000 points, 20 clusters and 20 features"""
nruns = 20
npoints = 10000
nclusters = 20
nfeatures = 20
points, clusters, partition = self.createEmptyParameterArrays(npoints, nclusters, nfeatures)
partition[:] = np.random.randint(0, nclusters, npoints)
for i in xrange(nclusters):
mask = partition == i
mean = np.random.uniform(-1.0, 1.0, nfeatures)
sigma = np.diag(np.random.uniform(0.1, 0.5, size=nfeatures))
points[mask] = np.random.multivariate_normal(mean, sigma, int(np.sum(mask)))
l = np.random.randint(0, npoints, nclusters)
clusters = points[l].copy()
partition[:] = np.random.randint(0, nclusters, npoints)
backup_partition = partition.copy()
backup_clusters = clusters.copy()
first_partition = None
first_clusters = None
for i in xrange(nruns):
minkowski_p = 2
swap_threshold = 0
ccluster.kmeans(points, clusters, partition, minkowski_p, swap_threshold)
if first_partition is None:
first_partition = partition.copy()
first_clusters = clusters.copy()
else:
partition_equal_first_partition = (partition == first_partition).all()
self.assertTrue(partition_equal_first_partition)
clusters_equal_first_clusters = (clusters == first_clusters).all()
self.assertTrue(clusters_equal_first_clusters)
partition = backup_partition.copy()
clusters = backup_clusters.copy()
def testSimpleRun(self):
"""testing a single run with 10 points, 2 clusters and 1 feature"""
points, clusters, partition = self.createEmptyParameterArrays(10, 2, 1)
points[:5] = np.random.normal(-1.0, 0.5, (5,1))
points[5:] = np.random.normal(1.0, 0.5, (5,1))
clusters[0,0] = -10.0
clusters[1,0] = -5.0
partition[:] = 0
minkowski_p = 2
swap_threshold = 0
ccluster.kmeans(points, clusters, partition, minkowski_p, swap_threshold)
def testSimpleRun(self):
"""testing a single run with 10 points, 2 clusters and 1 feature"""
points, clusters, partition = self.createEmptyParameterArrays(10, 2, 1)
points[:5] = np.random.normal(-1.0, 0.5, (5, 1))
points[5:] = np.random.normal(1.0, 0.5, (5, 1))
clusters[0, 0] = -10.0
clusters[1, 0] = -5.0
partition[:] = 0
minkowski_p = 2
swap_threshold = 0
ccluster.kmeans(points, clusters, partition, minkowski_p,
swap_threshold)
def testComplexRun(self):
"""testing a single run with 50000 points, 50 clusters and 50 features"""
npoints = 50000
nclusters = 50
nfeatures = 50
points, clusters, partition = self.createEmptyParameterArrays(npoints, nclusters, nfeatures)
partition[:] = np.random.randint(0, nclusters, npoints)
for i in xrange(nclusters):
mask = partition == i
mean = np.random.uniform(-1.0, 1.0, nfeatures)
sigma = np.diag(np.random.uniform(0.1, 0.5, nfeatures))
s = np.sum(mask)
points[mask,:] = np.random.multivariate_normal(mean, sigma, int(np.sum(mask)))
l = np.random.randint(0, npoints, nclusters)
clusters = points[l].copy()
partition[:] = np.random.randint(0, nclusters, npoints)
minkowski_p = 2
swap_threshold = 0
ccluster.kmeans(points, clusters, partition, minkowski_p, swap_threshold)
def testMultipleRuns(self):
"""testing 50 runs with 10000 points, 20 clusters and 20 features"""
nruns = 20
npoints = 10000
nclusters = 20
nfeatures = 20
points, clusters, partition = self.createEmptyParameterArrays(
npoints, nclusters, nfeatures)
partition[:] = np.random.randint(0, nclusters, npoints)
for i in xrange(nclusters):
mask = partition == i
mean = np.random.uniform(-1.0, 1.0, nfeatures)
sigma = np.diag(np.random.uniform(0.1, 0.5, size=nfeatures))
points[mask] = np.random.multivariate_normal(
mean, sigma, int(np.sum(mask)))
l = np.random.randint(0, npoints, nclusters)
clusters = points[l].copy()
partition[:] = np.random.randint(0, nclusters, npoints)
backup_partition = partition.copy()
backup_clusters = clusters.copy()
first_partition = None
first_clusters = None
for i in xrange(nruns):
minkowski_p = 2
swap_threshold = 0
ccluster.kmeans(points, clusters, partition, minkowski_p,
swap_threshold)
if first_partition is None:
first_partition = partition.copy()
first_clusters = clusters.copy()
else:
partition_equal_first_partition = (
partition == first_partition).all()
self.assertTrue(partition_equal_first_partition)
clusters_equal_first_clusters = (
clusters == first_clusters).all()
self.assertTrue(clusters_equal_first_clusters)
partition = backup_partition.copy()
clusters = backup_clusters.copy()
def testComplexRun(self):
"""testing a single run with 50000 points, 50 clusters and 50 features"""
npoints = 50000
nclusters = 50
nfeatures = 50
points, clusters, partition = self.createEmptyParameterArrays(
npoints, nclusters, nfeatures)
partition[:] = np.random.randint(0, nclusters, npoints)
for i in xrange(nclusters):
mask = partition == i
mean = np.random.uniform(-1.0, 1.0, nfeatures)
sigma = np.diag(np.random.uniform(0.1, 0.5, nfeatures))
s = np.sum(mask)
points[mask, :] = np.random.multivariate_normal(
mean, sigma, int(np.sum(mask)))
l = np.random.randint(0, npoints, nclusters)
clusters = points[l].copy()
partition[:] = np.random.randint(0, nclusters, npoints)
minkowski_p = 2
swap_threshold = 0
ccluster.kmeans(points, clusters, partition, minkowski_p,
swap_threshold)
