def calculate_accuracy(new_y, verification_y):
"""
Calculates the accuracy of classification/clustering-algorithms.
Note this only works with integer/discrete classes. For algorithms that give approximations an error function is
required.
Parameters
----------
new_y : ht.tensor of shape (n_samples, n_features), required
The new labels that where generated
verification_y : ht.tensor of shape (n_samples, n_features), required
Known labels
Returns
----------
float
the accuracy, number of properly labeled samples divided by amount of labels.
"""
if new_y.gshape != verification_y.gshape:
raise ValueError("Expecting results of same length, got {}, {}".format(
new_y.gshape, verification_y.gshape))
count = ht.sum(ht.where(new_y == verification_y, 1, 0))
return count / new_y.gshape[0]
def test_fit_iris_unsplit(self):
split = 0
# get some test data
iris = ht.load("heat/datasets/iris.csv", sep=";", split=split)
ht.random.seed(1)
# fit the clusters
k = 3
kmedoid = ht.cluster.KMedoids(n_clusters=k, random_state=1)
kmedoid.fit(iris)
# check whether the results are correct
self.assertIsInstance(kmedoid.cluster_centers_, ht.DNDarray)
self.assertEqual(kmedoid.cluster_centers_.shape, (k, iris.shape[1]))
# same test with init=kmedoids++
kmedoid = ht.cluster.KMedoids(n_clusters=k, init="kmedoids++")
kmedoid.fit(iris)
# check whether the results are correct
self.assertIsInstance(kmedoid.cluster_centers_, ht.DNDarray)
self.assertEqual(kmedoid.cluster_centers_.shape, (k, iris.shape[1]))
# check whether result is actually a datapoint
for i in range(kmedoid.cluster_centers_.shape[0]):
self.assertTrue(
ht.any(
ht.sum(ht.abs(kmedoid.cluster_centers_[i, :] - iris),
axis=1) == 0))
def __joint_log_likelihood(self, X):
"""
Adapted to HeAT from scikit-learn.
Calculates joint log-likelihood for n_samples to be assigned to each class.
Returns ht.DNDarray joint_log_likelihood(n_samples, n_classes).
"""
jll_size = self.classes_._DNDarray__array.numel()
jll_shape = (X.shape[0], jll_size)
joint_log_likelihood = ht.empty(jll_shape, dtype=X.dtype, split=X.split, device=X.device)
for i in range(jll_size):
jointi = ht.log(self.class_prior_[i])
n_ij = -0.5 * ht.sum(ht.log(2.0 * ht.pi * self.sigma_[i, :]))
n_ij -= 0.5 * ht.sum(((X - self.theta_[i, :]) ** 2) / (self.sigma_[i, :]), 1)
joint_log_likelihood[:, i] = jointi + n_ij
return joint_log_likelihood
def _normalized_symmetric_L(self, A):
degree = ht.sum(A, axis=1)
degree.resplit_(axis=None)
# Find stand-alone vertices with no connections
temp = torch.ones(degree.shape,
dtype=degree.larray.dtype,
device=degree.device.torch_device)
degree.larray = torch.where(degree.larray == 0, temp, degree.larray)
L = A / ht.sqrt(ht.expand_dims(degree, axis=1))
L = L / ht.sqrt(ht.expand_dims(degree, axis=0))
L = L * (-1.0)
L.fill_diagonal(1.0)
return L
def predict(self, X) -> ht.dndarray:
"""
Parameters
----------
X : ht.DNDarray
Input data to be predicted
"""
distances = ht.spatial.cdist(X, self.x)
_, indices = ht.topk(distances, self.num_neighbours, largest=False)
labels = self.y[indices.flatten()]
labels.balance_()
labels = ht.reshape(labels, (indices.gshape + (self.y.gshape[1], )))
labels = ht.sum(labels, axis=1)
maximums = ht.argmax(labels, axis=1)
return maximums
def predict(self, x: DNDarray) -> DNDarray:
"""
Predict the class labels for the provided data.
Parameters
----------
x : DNDarray
The test samples.
"""
distances = self.effective_metric_(x, self.x)
_, indices = ht.topk(distances, self.n_neighbors, largest=False)
predictions = self.y[indices.flatten()]
predictions.balance_()
predictions = ht.reshape(predictions,
(indices.gshape + (self.y.gshape[1], )))
predictions = ht.sum(predictions, axis=1)
self.classes_ = ht.argmax(predictions, axis=1)
return self.classes_
def test_spherical_clusters(self):
seed = 1
n = 20 * ht.MPI_WORLD.size
data = self.create_spherical_dataset(num_samples_cluster=n,
radius=1.0,
offset=4.0,
dtype=ht.float32,
random_state=seed)
kmedoid = ht.cluster.KMedoids(n_clusters=4, init="kmedoids++")
kmedoid.fit(data)
self.assertIsInstance(kmedoid.cluster_centers_, ht.DNDarray)
self.assertEqual(kmedoid.cluster_centers_.shape, (4, 3))
for i in range(kmedoid.cluster_centers_.shape[0]):
self.assertTrue(
ht.any(
ht.sum(ht.abs(kmedoid.cluster_centers_[i, :] - data),
axis=1) == 0))
# More Samples
n = 100 * ht.MPI_WORLD.size
data = self.create_spherical_dataset(num_samples_cluster=n,
radius=1.0,
offset=4.0,
dtype=ht.float32,
random_state=seed)
kmedoid = ht.cluster.KMedoids(n_clusters=4, init="kmedoids++")
kmedoid.fit(data)
self.assertIsInstance(kmedoid.cluster_centers_, ht.DNDarray)
self.assertEqual(kmedoid.cluster_centers_.shape, (4, 3))
# check whether result is actually a datapoint
for i in range(kmedoid.cluster_centers_.shape[0]):
self.assertTrue(
ht.any(
ht.sum(ht.abs(kmedoid.cluster_centers_[i, :] - data),
axis=1) == 0))
# different datatype
n = 20 * ht.MPI_WORLD.size
data = self.create_spherical_dataset(num_samples_cluster=n,
radius=1.0,
offset=4.0,
dtype=ht.float64,
random_state=seed)
kmedoid = ht.cluster.KMedoids(n_clusters=4, init="kmedoids++")
kmedoid.fit(data)
self.assertIsInstance(kmedoid.cluster_centers_, ht.DNDarray)
self.assertEqual(kmedoid.cluster_centers_.shape, (4, 3))
for i in range(kmedoid.cluster_centers_.shape[0]):
self.assertTrue(
ht.any(
ht.sum(ht.abs(kmedoid.cluster_centers_[i, :] -
data.astype(ht.float32)),
axis=1) == 0))
# on Ints (different radius, offset and datatype
data = self.create_spherical_dataset(num_samples_cluster=n,
radius=10.0,
offset=40.0,
dtype=ht.int32,
random_state=seed)
kmedoid = ht.cluster.KMedoids(n_clusters=4, init="kmedoids++")
kmedoid.fit(data)
self.assertIsInstance(kmedoid.cluster_centers_, ht.DNDarray)
self.assertEqual(kmedoid.cluster_centers_.shape, (4, 3))
for i in range(kmedoid.cluster_centers_.shape[0]):
self.assertTrue(
ht.any(
ht.sum(ht.abs(kmedoid.cluster_centers_[i, :] - data),
axis=1) == 0))
def test_sum(self):
array_len = 11
# check sum over all float elements of 1d tensor locally
shape_noaxis = ht.ones(array_len)
no_axis_sum = shape_noaxis.sum()
self.assertIsInstance(no_axis_sum, ht.DNDarray)
self.assertEqual(no_axis_sum.shape, (1, ))
self.assertEqual(no_axis_sum.lshape, (1, ))
self.assertEqual(no_axis_sum.dtype, ht.float32)
self.assertEqual(no_axis_sum.larray.dtype, torch.float32)
self.assertEqual(no_axis_sum.split, None)
self.assertEqual(no_axis_sum.larray, array_len)
out_noaxis = ht.zeros((1, ))
ht.sum(shape_noaxis, out=out_noaxis)
self.assertTrue(out_noaxis.larray == shape_noaxis.larray.sum())
# check sum over all float elements of split 1d tensor
shape_noaxis_split = ht.arange(array_len, split=0)
shape_noaxis_split_sum = shape_noaxis_split.sum()
self.assertIsInstance(shape_noaxis_split_sum, ht.DNDarray)
self.assertEqual(shape_noaxis_split_sum.shape, (1, ))
self.assertEqual(shape_noaxis_split_sum.lshape, (1, ))
self.assertEqual(shape_noaxis_split_sum.dtype, ht.int64)
self.assertEqual(shape_noaxis_split_sum.larray.dtype, torch.int64)
self.assertEqual(shape_noaxis_split_sum.split, None)
self.assertEqual(shape_noaxis_split_sum, 55)
out_noaxis = ht.zeros((1, ))
ht.sum(shape_noaxis_split, out=out_noaxis)
self.assertEqual(out_noaxis.larray, 55)
# check sum over all float elements of 3d tensor locally
shape_noaxis = ht.ones((3, 3, 3))
no_axis_sum = shape_noaxis.sum()
self.assertIsInstance(no_axis_sum, ht.DNDarray)
self.assertEqual(no_axis_sum.shape, (1, ))
self.assertEqual(no_axis_sum.lshape, (1, ))
self.assertEqual(no_axis_sum.dtype, ht.float32)
self.assertEqual(no_axis_sum.larray.dtype, torch.float32)
self.assertEqual(no_axis_sum.split, None)
self.assertEqual(no_axis_sum.larray, 27)
out_noaxis = ht.zeros((1, ))
ht.sum(shape_noaxis, out=out_noaxis)
self.assertEqual(out_noaxis.larray, 27)
# check sum over all float elements of split 3d tensor
shape_noaxis_split_axis = ht.ones((3, 3, 3), split=0)
split_axis_sum = shape_noaxis_split_axis.sum(axis=0)
self.assertIsInstance(split_axis_sum, ht.DNDarray)
self.assertEqual(split_axis_sum.shape, (3, 3))
self.assertEqual(split_axis_sum.dtype, ht.float32)
self.assertEqual(split_axis_sum.larray.dtype, torch.float32)
self.assertEqual(split_axis_sum.split, None)
# check split semantics
shape_noaxis_split_axis = ht.ones((3, 3, 3), split=2)
split_axis_sum = shape_noaxis_split_axis.sum(axis=1)
self.assertIsInstance(split_axis_sum, ht.DNDarray)
self.assertEqual(split_axis_sum.shape, (3, 3))
self.assertEqual(split_axis_sum.dtype, ht.float32)
self.assertEqual(split_axis_sum.larray.dtype, torch.float32)
self.assertEqual(split_axis_sum.split, 1)
out_noaxis = ht.zeros((3, 3))
ht.sum(shape_noaxis, axis=0, out=out_noaxis)
self.assertTrue((out_noaxis.larray == torch.full(
(3, 3), 3, dtype=torch.float,
device=self.device.torch_device)).all())
# check sum over all float elements of splitted 5d tensor with negative axis
shape_noaxis_split_axis_neg = ht.ones((1, 2, 3, 4, 5), split=1)
shape_noaxis_split_axis_neg_sum = shape_noaxis_split_axis_neg.sum(
axis=-2)
self.assertIsInstance(shape_noaxis_split_axis_neg_sum, ht.DNDarray)
self.assertEqual(shape_noaxis_split_axis_neg_sum.shape, (1, 2, 3, 5))
self.assertEqual(shape_noaxis_split_axis_neg_sum.dtype, ht.float32)
self.assertEqual(shape_noaxis_split_axis_neg_sum.larray.dtype,
torch.float32)
self.assertEqual(shape_noaxis_split_axis_neg_sum.split, 1)
out_noaxis = ht.zeros((1, 2, 3, 5), split=1)
ht.sum(shape_noaxis_split_axis_neg, axis=-2, out=out_noaxis)
# check sum over all float elements of splitted 3d tensor with tuple axis
shape_split_axis_tuple = ht.ones((3, 4, 5), split=1)
shape_split_axis_tuple_sum = shape_split_axis_tuple.sum(axis=(-2, -3))
expected_result = ht.ones((5, )) * 12.0
self.assertIsInstance(shape_split_axis_tuple_sum, ht.DNDarray)
self.assertEqual(shape_split_axis_tuple_sum.shape, (5, ))
self.assertEqual(shape_split_axis_tuple_sum.dtype, ht.float32)
self.assertEqual(shape_split_axis_tuple_sum.larray.dtype,
torch.float32)
self.assertEqual(shape_split_axis_tuple_sum.split, None)
self.assertTrue((shape_split_axis_tuple_sum == expected_result).all())
# exceptions
with self.assertRaises(ValueError):
ht.ones(array_len).sum(axis=1)
with self.assertRaises(ValueError):
ht.ones(array_len).sum(axis=-2)
with self.assertRaises(ValueError):
ht.ones((4, 4)).sum(axis=0, out=out_noaxis)
with self.assertRaises(TypeError):
ht.ones(array_len).sum(axis="bad_axis_type")
def logsumexp(self,
a,
axis=None,
b=None,
keepdim=False,
return_sign=False):
"""
Adapted to HeAT from scikit-learn.
Compute the log of the sum of exponentials of input elements.
Parameters
----------
a : ht.tensor
Input array.
axis : None or int or tuple of ints, optional
Axis or axes over which the sum is taken. By default `axis` is None,
and all elements are summed.
keepdim : bool, optional
If this is set to True, the axes which are reduced are left in the
result as dimensions with size one. With this option, the result
will broadcast correctly against the original array.
b : ht.tensor, optional
Scaling factor for exp(`a`) must be of the same shape as `a` or
broadcastable to `a`. These values may be negative in order to
implement subtraction.
#return_sign : bool, optional
If this is set to True, the result will be a pair containing sign
information; if False, results that are negative will be returned
as NaN. Default is False (no sign information).
#TODO: returns NotImplementedYet error.
Returns
-------
res : ht.tensor
The result, ``np.log(np.sum(np.exp(a)))`` calculated in a numerically
more stable way. If `b` is given then ``np.log(np.sum(b*np.exp(a)))``
is returned.
#TODO sgn : ndarray NOT IMPLEMENTED YET
If return_sign is True, this will be an array of floating-point
numbers matching res and +1, 0, or -1 depending on the sign
of the result. If False, only one result is returned.
"""
if b is not None:
raise NotImplementedError("Not implemented for weighted logsumexp")
a_max = ht.max(a, axis=axis, keepdim=True)
# TODO: sanitize a_max / implement isfinite(): sanitation module, cf. #468
# if a_max.numdims > 0:
# a_max[~np.isfinite(a_max)] = 0
# elif not np.isfinite(a_max):
# a_max = 0
# TODO: reinstate after allowing b not None
# if b is not None:
# b = np.asarray(b)
# tmp = b * np.exp(a - a_max)
# else:
tmp = ht.exp(a - a_max)
s = ht.sum(tmp, axis=axis, keepdim=keepdim)
if return_sign:
raise NotImplementedError("Not implemented for return_sign")
# sgn = np.sign(s) # TODO: np.sign
# s *= sgn # /= makes more sense but we need zero -> zero
out = ht.log(s)
if not keepdim:
a_max = ht.squeeze(a_max, axis=axis)
out += a_max
# if return_sign: #TODO: np.sign
# return out, sgn
# else:
return out
def _simple_L(self, A):
degree = ht.sum(A, axis=1)
L = ht.diag(degree) - A
return L
def test_sum(self):
array_len = 11
# check sum over all float elements of 1d tensor locally
shape_noaxis = ht.ones(array_len)
no_axis_sum = shape_noaxis.sum()
self.assertIsInstance(no_axis_sum, ht.tensor)
self.assertEqual(no_axis_sum.shape, (1, ))
self.assertEqual(no_axis_sum.lshape, (1, ))
self.assertEqual(no_axis_sum.dtype, ht.float32)
self.assertEqual(no_axis_sum._tensor__array.dtype, torch.float32)
self.assertEqual(no_axis_sum.split, None)
self.assertEqual(no_axis_sum._tensor__array, array_len)
out_noaxis = ht.zeros((1, ))
ht.sum(shape_noaxis, out=out_noaxis)
self.assertTrue(
out_noaxis._tensor__array == shape_noaxis._tensor__array.sum())
# check sum over all float elements of split 1d tensor
shape_noaxis_split = ht.arange(array_len, split=0)
shape_noaxis_split_sum = shape_noaxis_split.sum()
self.assertIsInstance(shape_noaxis_split_sum, ht.tensor)
self.assertEqual(shape_noaxis_split_sum.shape, (1, ))
self.assertEqual(shape_noaxis_split_sum.lshape, (1, ))
self.assertEqual(shape_noaxis_split_sum.dtype, ht.int64)
self.assertEqual(shape_noaxis_split_sum._tensor__array.dtype,
torch.int64)
self.assertEqual(shape_noaxis_split_sum.split, None)
self.assertEqual(shape_noaxis_split_sum, 55)
out_noaxis = ht.zeros((1, ))
ht.sum(shape_noaxis_split, out=out_noaxis)
self.assertEqual(out_noaxis._tensor__array, 55)
# check sum over all float elements of 3d tensor locally
shape_noaxis = ht.ones((3, 3, 3))
no_axis_sum = shape_noaxis.sum()
self.assertIsInstance(no_axis_sum, ht.tensor)
self.assertEqual(no_axis_sum.shape, (1, ))
self.assertEqual(no_axis_sum.lshape, (1, ))
self.assertEqual(no_axis_sum.dtype, ht.float32)
self.assertEqual(no_axis_sum._tensor__array.dtype, torch.float32)
self.assertEqual(no_axis_sum.split, None)
self.assertEqual(no_axis_sum._tensor__array, 27)
out_noaxis = ht.zeros((1, ))
ht.sum(shape_noaxis, out=out_noaxis)
self.assertEqual(out_noaxis._tensor__array, 27)
# check sum over all float elements of split 3d tensor
shape_noaxis_split_axis = ht.ones((3, 3, 3), split=0)
split_axis_sum = shape_noaxis_split_axis.sum(axis=0)
self.assertIsInstance(split_axis_sum, ht.tensor)
self.assertEqual(split_axis_sum.shape, (1, 3, 3))
self.assertEqual(split_axis_sum.dtype, ht.float32)
self.assertEqual(split_axis_sum._tensor__array.dtype, torch.float32)
self.assertEqual(split_axis_sum.split, None)
out_noaxis = ht.zeros((
1,
3,
3,
))
ht.sum(shape_noaxis, axis=0, out=out_noaxis)
self.assertTrue((out_noaxis._tensor__array == torch.full((
1,
3,
3,
), 3)).all())
# check sum over all float elements of splitted 5d tensor with negative axis
shape_noaxis_split_axis_neg = ht.ones((1, 2, 3, 4, 5), split=1)
shape_noaxis_split_axis_neg_sum = shape_noaxis_split_axis_neg.sum(
axis=-2)
self.assertIsInstance(shape_noaxis_split_axis_neg_sum, ht.tensor)
self.assertEqual(shape_noaxis_split_axis_neg_sum.shape,
(1, 2, 3, 1, 5))
self.assertEqual(shape_noaxis_split_axis_neg_sum.dtype, ht.float32)
self.assertEqual(shape_noaxis_split_axis_neg_sum._tensor__array.dtype,
torch.float32)
self.assertEqual(shape_noaxis_split_axis_neg_sum.split, 1)
out_noaxis = ht.zeros((1, 2, 3, 1, 5))
ht.sum(shape_noaxis_split_axis_neg, axis=-2, out=out_noaxis)
# exceptions
with self.assertRaises(ValueError):
ht.ones(array_len).sum(axis=1)
with self.assertRaises(ValueError):
ht.ones(array_len).sum(axis=-2)
with self.assertRaises(ValueError):
ht.ones((4, 4)).sum(axis=0, out=out_noaxis)
with self.assertRaises(TypeError):
ht.ones(array_len).sum(axis='bad_axis_type')
