def get_distance_matrix(self, neighborhood_radius = None, copy = True):
"""
Parameters
----------
neighborhood_radius : scalar, passed to distance_matrix. Value such that all
distances beyond neighborhood_radius are considered infinite.
if this value is not passed the value of self.neighborhood_radius is used
copy : boolean, whether to return a copied version of the distance matrix
Returns
-------
self.distance_matrix : sparse Ndarray (N_obs, N_obs). Non explicit 0.0 values
(e.g. diagonal) should be considered Infinite.
"""
if self.input_type == 'affinity':
raise ValueError("input_method was passed as affinity. "
"Distance matrix cannot be computed.")
if self.distance_matrix is None:
# if there's no existing distance matrix we make one
if ((neighborhood_radius is not None) and (neighborhood_radius != self.neighborhood_radius)):
# a different radius was passed than self.neighborhood_radius
self.neighborhood_radius = neighborhood_radius
self.distance_matrix = distance_matrix(self.X, method = self.distance_method,
flindex = self.flindex,
radius = self.neighborhood_radius,
cyindex = self.cyindex)
else:
# if there is an existing matrix we have to see if we need to overwrite
if ((neighborhood_radius is not None) and (neighborhood_radius != self.neighborhood_radius)):
# if there's a new radius we need to re-calculate
if self.input_type == 'distance':
# but if we were passed distance this is impossible
raise ValueError("input_method was passed as distance."
"requested radius not equal to self.neighborhood_radius."
"distance matrix cannot be re-calculated.")
else:
# if we were passed data then we need to re-calculate
self.neighborhood_radius = neighborhood_radius
self.distance_matrix = distance_matrix(self.X, method = self.distance_method,
flindex = self.flindex,
radius = self.neighborhood_radius,
cyindex = self.cyindex)
if copy:
return self.distance_matrix.copy()
else:
return self.distance_matrix
def test_all_methods_close(almost_equal_decimals = 5):
flindex = None
try:
import pyflann as pyf
except ImportError:
try:
# use "export FLANN_ROOT=<FLANN_ROOT>"to set enviromental variable
path_to_flann = os.environ['FLANN_ROOT'] + '/src/python'
sys.path.insert(0, path_to_flann)
import pyflann as pyf
except ImportError:
warnings.warn("pyflann not installed. Will not test pyflann method")
t1 = time.clock()
D1 = distance_matrix(X, method = 'cython')
print "cython version:",time.clock() - t1
flindex = pyf.FLANN()
flparams = flindex.build_index(X, algorithm = 'kmeans', target_precision = 0.9)
t2 = time.clock()
D2 = distance_matrix(X, method = 'pyflann', flindex = flindex)
print "pyflann version:",time.clock() - t2
t3 = time.clock()
D3 = distance_matrix(X, method = 'cyflann')
print "cyflann version:",time.clock() - t3
t4 = time.clock()
D4 = distance_matrix(X, method = 'brute')
print "brute version:",time.clock() - t4
D5 = distance_matrix(X)
assert_array_almost_equal(D1.todense(), D, almost_equal_decimals)
assert_array_almost_equal(D2.todense(), D, almost_equal_decimals)
assert_array_almost_equal(D3.todense(), D, almost_equal_decimals)
assert_array_almost_equal(D4.todense(), D, almost_equal_decimals)
assert_array_almost_equal(D5.todense(), D, almost_equal_decimals)
def test_laplacian_create_A_sparse():
"""
Test that A_sparse is the same as A_dense for a small A matrix
"""
rad = 2.
n_samples = 6
X = np.arange(n_samples)
X = X[ :,np.newaxis]
X = np.concatenate((X,np.zeros((n_samples,1),dtype=float)),axis=1)
X = np.asarray( X, order="C" )
test_dist_matrix = distance_matrix( X, radius = rad )
A_dense = affinity_matrix( test_dist_matrix.toarray(), rad, symmetrize = False )
A_sparse = affinity_matrix( sparse.csr_matrix( test_dist_matrix ), rad, symmetrize = False )
A_spdense = A_sparse.toarray()
A_spdense[ A_spdense == 0 ] = 1.
print( 'A_dense')
print( A_dense )
print( 'A_sparse', A_spdense )
assert_array_equal( A_dense, A_spdense )
def test_get_laplacian_matrix(almost_equal_decimals = 5):
""" test different ways to call get_laplacian_matrix """
# 1. (No existing laplacian, no passed type, no self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
Geometry = geom.Geometry(A, input_type = 'affinity')
## Return default laplacian
lapl = Geometry.get_laplacian_matrix()
lapl2 = graph_laplacian(A)
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
# 2. (No existing laplacian, no passed type, existing self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
for laplacian_type in laplacian_types:
Geometry = geom.Geometry(A, input_type = 'affinity', laplacian_type = laplacian_type)
## Return self.type laplacian
lapl = Geometry.get_laplacian_matrix()
lapl2 = graph_laplacian(A, normed = laplacian_type)
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
# 3. (No existing laplacian, passed type, no self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
for laplacian_type in laplacian_types:
Geometry = geom.Geometry(A, input_type = 'affinity')
## Return passed type laplacian, set self.type to passed type
lapl = Geometry.get_laplacian_matrix(laplacian_type = laplacian_type)
lapl2 = graph_laplacian(A, normed = laplacian_type)
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
assert(Geometry.laplacian_type == laplacian_type)
# 4. (No existing laplacian, passed type equal to self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
for laplacian_type in laplacian_types:
Geometry = geom.Geometry(A, input_type = 'affinity',laplacian_type = laplacian_type)
## Return passed type laplacian
lapl = Geometry.get_laplacian_matrix(laplacian_type = laplacian_type)
lapl2 = graph_laplacian(A, normed = laplacian_type)
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
assert(Geometry.laplacian_type == laplacian_type)
# 5. (No existing laplacian, passed type not equal to self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
for laplacian_type in laplacian_types:
if laplacian_type == 'geometric':
existing_type = 'randomwalk'
else:
existing_type = 'geometric'
Geometry = geom.Geometry(A, input_type = 'affinity', laplacian_type = existing_type)
## Set type to passed type, return passed type laplacian
lapl = Geometry.get_laplacian_matrix(laplacian_type = laplacian_type)
lapl2 = graph_laplacian(A, normed = laplacian_type)
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
assert(Geometry.laplacian_type == laplacian_type)
# 6. (Existing laplacian, no passed type, no self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
Geometry = geom.Geometry(A, input_type = 'affinity')
lapl2 = graph_laplacian(A)
Geometry.assign_laplacian_matrix(lapl2)
## Return existing laplacian
lapl = Geometry.get_laplacian_matrix()
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
# 7. (Existing laplacian, no passed type, existing self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
Geometry = geom.Geometry(A, input_type = 'affinity', laplacian_type = 'geometric')
lapl2 = graph_laplacian(A)
Geometry.assign_laplacian_matrix(lapl2)
## Return existing laplacian
lapl = Geometry.get_laplacian_matrix()
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
# 8. (Existing laplacian, passed type, no self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
Geometry = geom.Geometry(A, input_type = 'affinity')
lapl2 = graph_laplacian(A)
Geometry.assign_laplacian_matrix(lapl2)
## Calculate passed type, set self.type to passed type
# this will warn that it will overwrite:
with warnings.catch_warnings():
warnings.simplefilter("ignore")
lapl = Geometry.get_laplacian_matrix(laplacian_type = 'randomwalk')
lapl2 = graph_laplacian(A, normed = 'randomwalk')
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
# 9. (Existing laplacian, passed type equal to self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
for laplacian_type in laplacian_types:
Geometry = geom.Geometry(A, input_type = 'affinity', laplacian_type = laplacian_type)
lapl2 = graph_laplacian(A, normed = laplacian_type)
Geometry.assign_laplacian_matrix(lapl2, laplacian_type = laplacian_type)
## Return existing laplacian
lapl = Geometry.get_laplacian_matrix(laplacian_type = laplacian_type)
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
assert(Geometry.laplacian_type == laplacian_type)
# 10. (Existing laplacian, passed type not equal to self.type)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1)
A = affinity_matrix(dist_mat, neighbors_radius = 1)
for laplacian_type in laplacian_types:
if laplacian_type == 'geometric':
existing_type = 'randomwalk'
else:
existing_type = 'geometric'
Geometry = geom.Geometry(A, input_type = 'affinity', laplacian_type = existing_type)
lapl2 = graph_laplacian(A, normed = existing_type)
Geometry.assign_laplacian_matrix(lapl2, laplacian_type = existing_type)
## Calculate passed type, set self.type to passed type
# this will warn that it will overwrite:
with warnings.catch_warnings():
warnings.simplefilter("ignore")
lapl = Geometry.get_laplacian_matrix(laplacian_type = laplacian_type)
lapl2 = graph_laplacian(A, normed = laplacian_type)
assert_array_almost_equal(lapl.todense(), lapl2.todense(),almost_equal_decimals)
assert(Geometry.laplacian_type == laplacian_type)
def test_Geometry_distance(almost_equal_decimals = 5):
geom = Geometry(X)
D1 = geom.get_distance_matrix()
D2 = distance_matrix(X)
assert_array_almost_equal(D1.todense(), D2.todense(), almost_equal_decimals)
def test_get_affinity_matrix(almost_equal_decimals=5):
""" test different ways to call get_affinity_matrix """
# 1. (No existing affinity, no passed radius, no self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1/X.shape[1])
A2 = affinity_matrix(dist_mat, neighbors_radius = 1/X.shape[1])
Geometry = geom.Geometry(X, input_type = 'data')
## Return default Affinity
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 2. (No existing affinity, no passed radius, no self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
dist_mat = distance_matrix(X, radius = 1/X.shape[1])
A2 = affinity_matrix(dist_mat, neighbors_radius = 1/X.shape[1])
Geometry = geom.Geometry(dist_mat, input_type = 'distance')
## Return default
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 3. (No existing affinity, no passed radius, existing self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = radius,
affinity_radius = radius)
## Return with existing radius
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 4. (No existing affinity, no passed radius, existing self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(dist_mat, input_type = 'distance', neighborhood_radius = radius,
affinity_radius = radius)
## Return with existing radius
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 5. (No existing affinity, passed radius, no existing self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = 2)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = radius)
## Return passed radius affinity, set radius to passed
A = Geometry.get_affinity_matrix(affinity_radius=2)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 6. (No existing affinity, passed radius, no existing self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = 2)
Geometry = geom.Geometry(dist_mat, input_type = 'distance', neighborhood_radius = radius)
## Return passed radius affinity, set radius to passed
A = Geometry.get_affinity_matrix(affinity_radius=2)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 7. (No existing affinity, passed radius equal to existing self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = radius,
affinity_radius = radius)
## Return self.radius affinity
A = Geometry.get_affinity_matrix(affinity_radius=radius)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 8. (No existing affinity, passed radius equal to existing self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(dist_mat, input_type = 'distance', neighborhood_radius = radius,
affinity_radius = radius)
## Return self.radius affinity
A = Geometry.get_affinity_matrix(affinity_radius=radius)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 9. (No existing affinity, passed radius not equal to existing self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
radius = 1
radius2 = 2
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius2)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = radius,
affinity_radius = radius)
## Return passed radius affinity, set radius to passed
A = Geometry.get_affinity_matrix(affinity_radius=radius2)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
assert(Geometry.affinity_radius == radius2)
# 10. (No existing affinity, passed radius not equal to existing self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
radius = 1
radius2 = 2
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius2)
Geometry = geom.Geometry(dist_mat, input_type = 'distance', neighborhood_radius = radius,
affinity_radius = radius)
## Return passed radius affinity, set radius to passed
A = Geometry.get_affinity_matrix(affinity_radius=radius2)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
assert(Geometry.affinity_radius == radius2)
# 11. (Existing affinity, no passed radius, no self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(X, input_type = 'data')
Geometry.assign_affinity_matrix(A2)
## Return existing affinity
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 12. (Existing affinity, no passed radius, no self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(dist_mat, input_type = 'distance')
Geometry.assign_affinity_matrix(A2)
## Return existing affinity
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 13. (Existing affinity, no passed radius, no self.radius, input_type = affinity)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(A2, input_type = 'affinity')
### Return existing affinity
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 14. (Existing affinity, no passed radius, existing self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = radius,
affinity_radius = radius)
Geometry.assign_affinity_matrix(A2)
### Return existing affinity
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 15. (Existing affinity, no passed radius, existing self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(dist_mat, input_type = 'distance', neighborhood_radius = radius,
affinity_radius = radius)
Geometry.assign_affinity_matrix(A2)
### Return existing affinity
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 16. (Existing affinity, no passed radius, existing self.radius, input_type = affinity)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(A2, input_type = 'affinity', neighborhood_radius = radius,
affinity_radius = radius)
## Return existing affinity
A = Geometry.get_affinity_matrix()
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 17. (Existing affinity, passed radius, no self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = 2)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = radius)
Geometry.assign_affinity_matrix(A2)
## Compute with passed radius, set self.radius to passed radius
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
A = Geometry.get_affinity_matrix(affinity_radius=radius)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 18. (Existing affinity, passed radius, no self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = 2)
Geometry = geom.Geometry(dist_mat, input_type = 'distance', neighborhood_radius = radius)
Geometry.assign_affinity_matrix(A2)
## Compute with passed radius, set self.radius to passed radius
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
A = Geometry.get_affinity_matrix(affinity_radius=radius)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 19. (Existing affinity, passed radius, no self.radius, input_type = affinity)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = 2)
Geometry = geom.Geometry(A2, input_type = 'affinity', neighborhood_radius = radius)
## Raise error, unknown existing radius, passed radius but input type affinity
msg = ("Input_method was passed as affinity."
"Requested radius was not equal to self.affinity_radius."
"Affinity Matrix cannot be recalculated.")
assert_raise_message(ValueError, msg, Geometry.get_affinity_matrix,
affinity_radius=radius)
# 20. (Existing affinity, passed radius equal to self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = radius,
affinity_radius = radius)
Geometry.assign_affinity_matrix(A2)
## Return existing affinity
A = Geometry.get_affinity_matrix(affinity_radius=radius)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 21. (Existing affinity, passed radius equal to self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(dist_mat, input_type = 'distance', neighborhood_radius = radius,
affinity_radius = radius)
Geometry.assign_affinity_matrix(A2)
## Return existing affinity
A = Geometry.get_affinity_matrix(affinity_radius=radius)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 22. (Existing affinity, passed radius equal to self.radius, input_type = affinity)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(A2, input_type = 'affinity', neighborhood_radius = radius,
affinity_radius = radius)
## Return existing affinity
A = Geometry.get_affinity_matrix(affinity_radius=radius)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
# 23. (Existing affinity, passed radius not equal to self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = radius,
affinity_radius = radius)
Geometry.assign_affinity_matrix(A2)
## Re-calculate with passed radius, set self.radius to passed radius
A2 = affinity_matrix(dist_mat, neighbors_radius = 2)
A = Geometry.get_affinity_matrix(affinity_radius=2)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
assert(Geometry.affinity_radius == 2)
# 24. (Existing affinity, passed radius not equal to self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = radius)
Geometry = geom.Geometry(dist_mat, input_type = 'distance', neighborhood_radius = radius,
affinity_radius = radius)
Geometry.assign_affinity_matrix(A2)
## Re-calculate with passed radius, set self.radius to passed radius
A2 = affinity_matrix(dist_mat, neighbors_radius=2)
A = Geometry.get_affinity_matrix(affinity_radius=2)
assert_array_almost_equal(A.todense(), A2.todense(), almost_equal_decimals)
assert(Geometry.affinity_radius == 2)
# 25. (Existing affinity, passed radius not equal to self.radius, input_type = affinity)
X = np.random.uniform(size = (10,10))
radius = 1
dist_mat = distance_matrix(X, radius = radius)
A2 = affinity_matrix(dist_mat, neighbors_radius = 2)
Geometry = geom.Geometry(A2, input_type = 'affinity', neighborhood_radius = radius,
affinity_radius = 2)
## Raise error, passed affinity re-calculateion not supported.
msg = ("Input_method was passed as affinity."
"Requested radius was not equal to self.affinity_radius."
"Affinity Matrix cannot be recalculated.")
assert_raise_message(ValueError, msg, Geometry.get_affinity_matrix,
affinity_radius=radius)
def test_get_distance_matrix(almost_equal_decimals = 5):
""" test different ways to call get_distance_matrix """
# 1. (Input type = Affinity)
X = np.random.uniform(size = (10,10))
Geometry = geom.Geometry(X, input_type = 'affinity')
## Raise error Affinity matrix passed
msg = ( "input_method was passed as affinity. " "Distance matrix cannot be computed.")
assert_raise_message(ValueError, msg, Geometry.get_distance_matrix)
# 2. (No existing distance, no passed radius, no self.radius, input_type = Data)
X = np.random.uniform(size = (10,10))
Geometry = geom.Geometry(X, input_type = 'data')
## Calculate with default radius
distance_mat = Geometry.get_distance_matrix()
distance_mat2 = distance_matrix(X, radius = 1/X.shape[1])
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
# 3. (No existing distance, no passed radius, existing self.radius, input_type = Data)
X = np.random.uniform(size = (10,10))
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = 1)
## Calculate with existing radius
distance_mat = Geometry.get_distance_matrix()
distance_mat2 = distance_matrix(X, radius = 1)
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
# 4. (No existing distance, passed radius, no existing self.radius, input_type = Data)
X = np.random.uniform(size = (10,10))
Geometry = geom.Geometry(X, input_type = 'data')
## Set current radius to passed, calcualte with passed radius
distance_mat = Geometry.get_distance_matrix(neighborhood_radius = 1)
distance_mat2 = distance_matrix(X, radius = 1)
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
assert(Geometry.neighborhood_radius == 1)
# 5. (No existing distance, passed radius equal to existing self.radius, input_type = Data)
X = np.random.uniform(size = (10,10))
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = 1)
## Calculate with passed radius
distance_mat = Geometry.get_distance_matrix(neighborhood_radius = 1)
distance_mat2 = distance_matrix(X, radius = 1)
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
# 6. (No existing distance, passed radius not equal to existing self.radius, input_type = Data)
X = np.random.uniform(size = (10,10))
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = 1)
## Calculate with passed radius, set new self.radius
distance_mat = Geometry.get_distance_matrix(neighborhood_radius = 2)
distance_mat2 = distance_matrix(X, radius = 2)
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
assert(Geometry.neighborhood_radius == 2)
# 7. (Existing distance, no passed radius, no self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
Geometry = geom.Geometry(X, input_type = 'data')
distance_mat2 = distance_matrix(X, radius = 1)
Geometry.assign_distance_matrix(distance_mat2)
## Return existing distance matrix
distance_mat = Geometry.get_distance_matrix()
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
# 8. (Existing distance, no passed radius, no self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
distance_mat2 = distance_matrix(X, radius = 1)
Geometry = geom.Geometry(distance_mat2, input_type = 'distance')
## Return existing distance matrix
distance_mat = Geometry.get_distance_matrix()
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
# 9. (Existing distance, no passed radius, existing self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius = 1)
distance_mat2 = distance_matrix(X, radius = 1)
Geometry.assign_distance_matrix(distance_mat2)
## Return existing distance Matrix
distance_mat = Geometry.get_distance_matrix()
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
# 10. (Existing distance, no passed radius, existing self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
distance_mat2 = distance_matrix(X, radius = 1)
Geometry = geom.Geometry(distance_mat2, input_type = 'distance', neighborhood_radius = 1)
## Return existing distance matrix
distance_mat = Geometry.get_distance_matrix()
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
# 11. (Existing distance, passed radius, no existing self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
distance_mat2 = distance_matrix(X, radius = 1)
Geometry = geom.Geometry(X, input_type = 'data')
Geometry.assign_distance_matrix(distance_mat2)
## Re-calculate with passed radius, set self.radius to passed radius
distance_mat = Geometry.get_distance_matrix(neighborhood_radius=3)
distance_mat2 = distance_matrix(X, radius = 3)
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
assert(Geometry.neighborhood_radius == 3)
# 12. (Existing distance, passed radius, no existing self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
distance_mat2 = distance_matrix(X, radius = 1)
Geometry = geom.Geometry(distance_mat2, input_type = 'distance')
## Raise error, can't re-calculate
msg = ("input_method was passed as distance."
"requested radius not equal to self.neighborhood_radius."
"distance matrix cannot be re-calculated.")
assert_raise_message(ValueError, msg, Geometry.get_distance_matrix, neighborhood_radius=3)
# 13. (Existing distance, passed radius equal to existing self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
distance_mat2 = distance_matrix(X, radius = 1)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius=1)
Geometry.assign_distance_matrix(distance_mat2)
## Return existing distance matrix
distance_mat = Geometry.get_distance_matrix(neighborhood_radius=1)
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
# 14. (Existing distance, passed radius equal to existing self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
distance_mat2 = distance_matrix(X, radius = 1)
Geometry = geom.Geometry(distance_mat2, input_type = 'distance', neighborhood_radius = 1)
## Return existing distance matrix
distance_mat = Geometry.get_distance_matrix(neighborhood_radius=1)
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
# 15. (Existing distance, passed radius not equal to existing self.radius, input_type = data)
X = np.random.uniform(size = (10,10))
distance_mat2 = distance_matrix(X, radius = 1)
Geometry = geom.Geometry(X, input_type = 'data', neighborhood_radius=1)
Geometry.assign_distance_matrix(distance_mat2)
## Recompute with passed radius, set self.radius to passed radius
distance_mat2 = distance_matrix(X, radius = 3)
distance_mat = Geometry.get_distance_matrix(neighborhood_radius=3)
assert_array_almost_equal(distance_mat.todense(), distance_mat2.todense(), almost_equal_decimals)
assert(Geometry.neighborhood_radius == 3)
# 16. (Existing distance, passed radius not equal to existing self.radius, input_type = distance)
X = np.random.uniform(size = (10,10))
distance_mat2 = distance_matrix(X, radius = 1)
Geometry = geom.Geometry(distance_mat2, input_type = 'distance', neighborhood_radius = 1)
## Raise error, can't recalculate
msg = ("input_method was passed as distance."
"requested radius not equal to self.neighborhood_radius."
"distance matrix cannot be re-calculated.")
assert_raise_message(ValueError, msg, Geometry.get_distance_matrix, neighborhood_radius=3)
def time_function(X, method, radius, d = None, random_state = None):
if method == 'distance_cy':
t0 = time.time()
dmat = distance_matrix(X, method = 'cyflann', radius = radius)
t1 = time.time()
return (t1 - t0)
if method == 'distance_py':
path_to_flann = '/homes/jmcq/flann-1.8.4-src/src/python'
sys.path.insert(0, path_to_flann)
import pyflann as pyf
flindex = pyf.FLANN()
flparams = flindex.build_index(X, algorithm = 'kmeans', target_precision = 0.9)
t0 = time.time()
dmat = distance_matrix(X, method = 'pyflann', radius = radius, flindex = flindex)
t1 = time.time()
return (t1 - t0)
if method == 'distance_brute':
t0 = time.time()
dmat = distance_matrix(X, method = 'brute', radius = radius)
t1 = time.time()
return (t1 - t0)
elif method == 'spectral':
from Mmani.embedding.spectral_embedding import SpectralEmbedding
t0 = time.time()
SE = SpectralEmbedding(n_components=d, eigen_solver='amg',random_state=random_state,
neighborhood_radius=radius, distance_method='cyflann',
input_type='data',laplacian_type='geometric')
embedding = SE.fit_transform(X)
t1 = time.time()
return (t1 - t0)
elif method == 'isomap':
from Mmani.embedding.isomap import Isomap
t0 = time.time()
try:
isomap = Isomap(n_components=d, eigen_solver='amg',random_state=random_state,
neighborhood_radius=radius, distance_method='cyflann',
input_type='data')
embedding = isomap.fit_transform(X)
except:
isomap = Isomap(n_components=d, eigen_solver='arpack',random_state=random_state,
neighborhood_radius=radius, distance_method='cyflann',
input_type='data')
embedding = isomap.fit_transform(X)
t1 = time.time()
return (t1 - t0)
elif method == 'ltsa':
from Mmani.embedding.ltsa import LTSA
t0 = time.time()
ltsa = LTSA(n_components=d, eigen_solver='amg',random_state=random_state,
neighborhood_radius=radius, distance_method='cyflann',
input_type='data')
embedding = ltsa.fit_transform(X)
t1 = time.time()
return (t1 - t0)
elif method == 'lle':
from Mmani.embedding.locally_linear import LocallyLinearEmbedding
t0 = time.time()
lle = LocallyLinearEmbedding(n_components=d, eigen_solver='amg',random_state=random_state,
neighborhood_radius=radius, distance_method='cyflann',
input_type='data')
embedding = lle.fit_transform(X)
t1 = time.time()
return (t1 - t0)
else:
raise ValueError('benchmark method: ' + str(method) + ' not found.')
return None
import numpy as np from scipy import sparse from sklearn.datasets.samples_generator import make_swiss_roll rng = np.random.RandomState(123) ns = 3000 X, t = make_swiss_roll( ns, noise = 0.0, random_state = rng) X = np.asarray( X, order="C" ) nf = 750 rad0 = 2.5 dim = 2 rad = rad0/ns**(1./(dim+6)) #check the scaling n_noisef = nf - 3 noise_rad_frac = 0.1 noiserad = rad/np.sqrt(n_noisef) * noise_rad_frac Xnoise = rng.rand(ns, n_noisef) * noiserad X = np.hstack((X, Xnoise)) rad = rad*(1+noise_rad_frac) # add a fraction for noisy dimensions from Mmani.geometry.distance import distance_matrix dmat = distance_matrix(X, method = 'cython', radius = rad)