def transform(self, X):
"""Transform data X according to the fitted model.
Parameters
----------
X : array-like or sparse matrix, shape=(n_samples, n_features)
Document word matrix.
Returns
-------
doc_topic_distr : shape=(n_samples, n_topics)
Unnormalized document topic distribution for X.
"""
X = self._check_inference(X, "HierarchicalDirichletProcess.transform")
n_jobs = _get_n_jobs(self.n_jobs)
verbose = max(0, self.verbose-1)
with Parallel(n_jobs=n_jobs, verbose=verbose) as parallel:
doc_topic_distr, _, _ = self._e_step(X,
cal_sstats=False,
cal_doc_distr=True,
cal_likelihood=False,
parallel=parallel)
return doc_topic_distr
def _e_step(self, X, cal_sstats, random_init, parallel=None):
"""E-step in EM update.
Parameters
----------
X : array-like or sparse matrix, shape=(n_samples, n_features)
Document word matrix.
cal_sstats : boolean
Parameter that indicate whether to calculate sufficient statistics
or not. Set ``cal_sstats`` to True when we need to run M-step.
random_init : boolean
Parameter that indicate whether to initialize document topic
distribution randomly in the E-step. Set it to True in training
steps.
parallel : joblib.Parallel (optional)
Pre-initialized instance of joblib.Parallel.
Returns
-------
(doc_topic_distr, suff_stats) :
`doc_topic_distr` is unnormalized topic distribution for each
document. In the literature, this is called `gamma`.
`suff_stats` is expected sufficient statistics for the M-step.
When `cal_sstats == False`, it will be None.
"""
# Run e-step in parallel
random_state = self.random_state_ if random_init else None
# TODO: make Parallel._effective_n_jobs public instead?
n_jobs = _get_n_jobs(self.n_jobs)
if parallel is None:
parallel = Parallel(n_jobs=n_jobs,
verbose=max(0, self.verbose - 1))
results = parallel(
delayed(_update_doc_distribution)
(X[idx_slice, :], self.exp_dirichlet_component_,
self.doc_topic_prior_, self.max_doc_update_iter,
self.mean_change_tol, cal_sstats, random_state)
for idx_slice in gen_even_slices(X.shape[0], n_jobs))
# merge result
doc_topics, sstats_list = zip(*results)
doc_topic_distr = np.vstack(doc_topics)
if cal_sstats:
# This step finishes computing the sufficient statistics for the
# M-step.
suff_stats = np.zeros(self.components_.shape)
for sstats in sstats_list:
suff_stats += sstats
suff_stats *= self.exp_dirichlet_component_
else:
suff_stats = None
return (doc_topic_distr, suff_stats)
def _partition_estimators(n_estimators, n_jobs): """Private function used to partition estimators between jobs.""" # Compute the number of jobs n_jobs = min(_get_n_jobs(n_jobs), n_estimators) # Partition estimators between jobs n_estimators_per_job = (n_estimators // n_jobs) * np.ones(n_jobs, dtype=np.int) n_estimators_per_job[:n_estimators % n_jobs] += 1 starts = np.cumsum(n_estimators_per_job) return n_jobs, n_estimators_per_job.tolist(), [0] + starts.tolist()
def _partition_estimators(n_estimators, n_jobs):
"""Private function used to partition estimators between jobs."""
# Compute the number of jobs
n_jobs = min(_get_n_jobs(n_jobs), n_estimators)
# Partition estimators between jobs
n_estimators_per_job = (n_estimators // n_jobs) * np.ones(n_jobs,
dtype=np.int)
n_estimators_per_job[:n_estimators % n_jobs] += 1
starts = np.cumsum(n_estimators_per_job)
return n_jobs, n_estimators_per_job.tolist(), [0] + starts.tolist()
def _approximate_bound(self, X):
"""Calculate approximate log-likelihood for the model
Parameters
----------
X : array-like or sparse matrix, shape=(n_samples, n_features)
Document word matrix.
Returns
-------
likelihood : float
approximate log-likelihood for variational parameters
"""
likelihood = 0.0
# calculate doc likelihood
n_jobs = _get_n_jobs(self.n_jobs)
verbose = max(0, self.verbose-1)
with Parallel(n_jobs=n_jobs, verbose=verbose) as parallel:
_, _, doc_likelihood = self._e_step(X,
cal_sstats=False,
cal_doc_distr=False,
cal_likelihood=True,
parallel=parallel)
likelihood += doc_likelihood
# E[log(p(beta|eta) - log(q(beta|lambda))]
# `beta` is Dirichlet distribution
lambda_ = self.lambda_
elog_beta_ = self.elog_beta_
n_features = lambda_.shape[1]
likelihood += np.sum((self.eta - lambda_) * elog_beta_)
likelihood += np.sum(gammaln(lambda_) - gammaln(self.eta))
likelihood += np.sum(gammaln(self.eta * n_features) -
gammaln(np.sum(lambda_, 1)))
# E[log(p(v_k|omega)) - log(q(v_k|a_k))]
# `v_k` is Beta distribution
v_k = self.v_stick_
likelihood += (v_k.shape[1] * np.log(self.omega))
v_k_col_sum = np.sum(v_k, 0)
dig_sum = psi(v_k_col_sum)
likelihood += np.sum(
(np.array([1.0, self.omega])[:, np.newaxis] - v_k) *
(psi(v_k) - dig_sum))
likelihood += np.sum(gammaln(v_k))
likelihood -= np.sum(gammaln(v_k_col_sum))
return likelihood
def get_neighbors(self, X=None, n_neighbors=None, return_distance=True):
n_neighbors = self.n_neighbors
query_is_train = False
X = check_array(X, accept_sparse='csr')
train_size = self._fit_X.shape[0]
n_samples, _ = X.shape
sample_range = np.arange(n_samples)[:, None]
n_jobs = _get_n_jobs(self.n_jobs)
result = Parallel(n_jobs, backend='threading')(
delayed(self._tree.query, check_pickle=False)(
X[s], n_neighbors, return_distance)
for s in gen_even_slices(X.shape[0], n_jobs)
)
if return_distance:
dist, neigh_ind = tuple(zip(*result))
result = np.vstack(dist), np.vstack(neigh_ind)
else:
result = np.vstack(result)
return result
def partial_fit(self, X, y=None):
"""Online VB with Mini-Batch update.
Parameters
----------
X : array-like or sparse matrix, shape=(n_samples, n_features)
Document word matrix.
y : Ignored.
Returns
-------
self
"""
self._check_params()
X = self._check_non_neg_array(X,
"LatentDirichletAllocation.partial_fit")
n_samples, n_features = X.shape
batch_size = self.batch_size
# initialize parameters or check
if not hasattr(self, 'components_'):
self._init_latent_vars(n_features)
if n_features != self.components_.shape[1]:
raise ValueError("The provided data has %d dimensions while "
"the model was trained with feature size %d." %
(n_features, self.components_.shape[1]))
n_jobs = _get_n_jobs(self.n_jobs)
with Parallel(n_jobs=n_jobs,
verbose=max(0, self.verbose - 1)) as parallel:
for idx_slice in gen_batches(n_samples, batch_size):
self._em_step(X[idx_slice, :],
total_samples=self.total_samples,
batch_update=False,
parallel=parallel)
return self
def fit(self, X, y=None):
"""Learn model for the data X
Parameters
----------
X : array-like or sparse matrix, shape=(n_samples, n_features)
Document word matrix.
Returns
-------
self
"""
self._check_params()
X = self._check_non_neg_array(
X, "HierarchicalDirichletProcess.fit")
self._init_global_latent_vars(*X.shape)
n_jobs = _get_n_jobs(self.n_jobs)
verbose = max(0, self.verbose - 1)
evaluate_every = self.evaluate_every
with Parallel(n_jobs=n_jobs, verbose=verbose) as parallel:
for i in xrange(self.max_iter):
# batch update
_, sstats, _ = self._e_step(X,
cal_sstats=True,
cal_doc_distr=False,
cal_likelihood=False,
parallel=parallel)
self._m_step(sstats, n_samples=X.shape[0], online_update=False)
# check perplexity
if evaluate_every > 0 and (i + 1) % evaluate_every == 0:
bound = self.score(X)
if self.verbose:
print('iteration: %d, ELOB: %.4f' % (i + 1, bound))
self.n_iter_ += 1
return self
def fit(self, X, y=None):
"""Learn model for the data X with variational Bayes method.
When `learning_method` is 'online', use mini-batch update.
Otherwise, use batch update.
Parameters
----------
X : array-like or sparse matrix, shape=(n_samples, n_features)
Document word matrix.
y : Ignored.
Returns
-------
self
"""
self._check_params()
X = self._check_non_neg_array(X, "LatentDirichletAllocation.fit")
n_samples, n_features = X.shape
max_iter = self.max_iter
evaluate_every = self.evaluate_every
learning_method = self.learning_method
if learning_method is None:
warnings.warn(
"The default value for 'learning_method' will be "
"changed from 'online' to 'batch' in the release "
"0.20. This warning was introduced in 0.18.",
DeprecationWarning)
learning_method = 'online'
batch_size = self.batch_size
# initialize parameters
self._init_latent_vars(n_features)
# change to perplexity later
last_bound = None
n_jobs = _get_n_jobs(self.n_jobs)
with Parallel(n_jobs=n_jobs,
verbose=max(0, self.verbose - 1)) as parallel:
for i in xrange(max_iter):
if learning_method == 'online':
for idx_slice in gen_batches(n_samples, batch_size):
self._em_step(X[idx_slice, :],
total_samples=n_samples,
batch_update=False,
parallel=parallel)
else:
# batch update
self._em_step(X,
total_samples=n_samples,
batch_update=True,
parallel=parallel)
# check perplexity
if evaluate_every > 0 and (i + 1) % evaluate_every == 0:
doc_topics_distr, _ = self._e_step(X,
cal_sstats=False,
random_init=False,
parallel=parallel)
bound = self._perplexity_precomp_distr(X,
doc_topics_distr,
sub_sampling=False)
if self.verbose:
print(
'iteration: %d of max_iter: %d, perplexity: %.4f' %
(i + 1, max_iter, bound))
if last_bound and abs(last_bound - bound) < self.perp_tol:
break
last_bound = bound
elif self.verbose:
print('iteration: %d of max_iter: %d' % (i + 1, max_iter))
self.n_iter_ += 1
# calculate final perplexity value on train set
doc_topics_distr, _ = self._e_step(X,
cal_sstats=False,
random_init=False,
parallel=parallel)
self.bound_ = self._perplexity_precomp_distr(X,
doc_topics_distr,
sub_sampling=False)
return self
def partial_fit(self, X, y=None):
"""Online VB with Mini-Batch update.
Parameters
----------
X : array-like or sparse matrix, shape=(n_samples, n_features)
Document word matrix.
Returns
-------
self
"""
self._check_params()
X = self._check_non_neg_array(X,
"LatentDirichletAllocation.partial_fit")
n_samples, n_features = X.shape
batch_size = self.batch_size
self.total_samples += n_samples
# initialize parameters or check
if not hasattr(self, 'components_'):
raise ValueError
# self._init_latent_vars(n_features)
if n_features != self.components_.shape[1]:
raise ValueError("The provided data has %d dimensions while "
"the model was trained with feature size %d." %
(n_features, self.components_.shape[1]))
n_jobs = _get_n_jobs(self.n_jobs)
max_iter = self.partial_max_iter
evaluate_every = self.partial_evaluate_every
self.n_partial_iter_ = 0
last_bound = None
doc_topic_distr = None
with Parallel(n_jobs=n_jobs,
verbose=max(0, self.verbose - 1)) as parallel:
for i in xrange(max_iter):
for idx_slice in gen_batches(n_samples, batch_size):
self._em_step(X[idx_slice, :],
total_samples=self.total_samples,
batch_update=False,
parallel=parallel)
# check perplexity
if evaluate_every > 0 and (i + 1) % evaluate_every == 0:
doc_topic_distr, _ = self._e_step(X,
cal_sstats=False,
random_init=False,
parallel=parallel)
bound = self.perplexity(X,
doc_topic_distr,
sub_sampling=False)
if self.verbose:
print('iteration: %d, perplexity: %.4f' %
(i + 1, bound))
if last_bound and abs(last_bound - bound) < self.perp_tol:
break
last_bound = bound
self.n_partial_iter_ += 1
if doc_topic_distr is None:
doc_topic_distr = self.transform(X)
else:
doc_topic_distr /= doc_topic_distr.sum(axis=1)[:, np.newaxis]
return doc_topic_distr
def kneighbors(self,
X=None,
E=None,
n_neighbors=None,
return_distance=True): #IY modified to account std dev
"""Finds the K-neighbors of a point.
Returns indices of and distances to the neighbors of each point.
Parameters
----------
X : array-like, shape (n_query, n_features), \
or (n_query, n_indexed) if metric == 'precomputed'
The query point or points.
If not provided, neighbors of each indexed point are returned.
In this case, the query point is not considered its own neighbor.
n_neighbors : int
Number of neighbors to get (default is the value
passed to the constructor).
return_distance : boolean, optional. Defaults to True.
If False, distances will not be returned
Returns
-------
dist : array
Array representing the lengths to points, only present if
return_distance=True
ind : array
Indices of the nearest points in the population matrix.
Examples
--------
In the following example, we construct a NeighborsClassifier
class from an array representing our data set and ask who's
the closest point to [1,1,1]
>>> samples = [[0., 0., 0.], [0., .5, 0.], [1., 1., .5]]
>>> from sklearn.neighbors import NearestNeighbors
>>> neigh = NearestNeighbors(n_neighbors=1)
>>> neigh.fit(samples) # doctest: +ELLIPSIS
NearestNeighbors(algorithm='auto', leaf_size=30, ...)
>>> print(neigh.kneighbors([[1., 1., 1.]])) # doctest: +ELLIPSIS
(array([[ 0.5]]), array([[2]]...))
As you can see, it returns [[0.5]], and [[2]], which means that the
element is at distance 0.5 and is the third element of samples
(indexes start at 0). You can also query for multiple points:
>>> X = [[0., 1., 0.], [1., 0., 1.]]
>>> neigh.kneighbors(X, return_distance=False) # doctest: +ELLIPSIS
array([[1],
[2]]...)
"""
if self._fit_method is None: #IY: usually in SOMPY is set to 'auto'
raise NotFittedError("Must fit neighbors before querying.")
if n_neighbors is None:
n_neighbors = self.n_neighbors
if X is not None:
query_is_train = False
X = check_array(X, accept_sparse='csr')
else:
query_is_train = True
X = self._fit_X
# Include an extra neighbor to account for the sample itself being
# returned, which is removed later. Results don't consider the test point itself as a neighbor
n_neighbors += 1
train_size = self._fit_X.shape[0]
if n_neighbors > train_size:
raise ValueError("Expected n_neighbors <= n_samples, "
" but n_samples = %d, n_neighbors = %d" %
(train_size, n_neighbors))
n_samples, _ = X.shape
sample_range = np.arange(n_samples)[:, None]
n_jobs = _get_n_jobs(self.n_jobs) #IY: single core at the moment
if self._fit_method == 'brute':
# for efficiency, use squared euclidean distances
if self.effective_metric_ == 'euclidean':
dist = pairwise_chidistances(
X,
self._fit_X,
'euclidean', #IY
n_jobs=n_jobs,
sigma=E,
squared=True)
else:
raise ValueError("kneighbor for project_realdata() works only"
" with euclidean metric")
neigh_ind = np.argpartition(dist, n_neighbors - 1, axis=1)
neigh_ind = neigh_ind[:, :n_neighbors]
# argpartition doesn't guarantee sorted order, so we sort again
neigh_ind = neigh_ind[sample_range,
np.argsort(dist[sample_range, neigh_ind])]
if return_distance:
result = np.sqrt(dist[sample_range, neigh_ind]), neigh_ind
else:
result = neigh_ind
## IY: only brute force at the moment...
#elif self._fit_method in ['ball_tree', 'kd_tree']:
# if issparse(X):
# raise ValueError(
# "%s does not work with sparse matrices. Densify the data, "
# "or set algorithm='brute'" % self._fit_method)
# result = Parallel(n_jobs, backend='threading')(
# delayed(self._tree.query, check_pickle=False)(
# X[s], n_neighbors, return_distance)
# for s in gen_even_slices(X.shape[0], n_jobs)
# )
# if return_distance:
# dist, neigh_ind = tuple(zip(*result))
# result = np.vstack(dist), np.vstack(neigh_ind)
# else:
# result = np.vstack(result)
else:
raise ValueError(
"only brute force algorithm accepted as _fit_method ")
if not query_is_train:
return result
else:
# If the query data is the same as the indexed data, we would like
# to ignore the first nearest neighbor of every sample, i.e
# the sample itself.
if return_distance:
dist, neigh_ind = result
else:
neigh_ind = result
sample_mask = neigh_ind != sample_range
# Corner case: When the number of duplicates are more
# than the number of neighbors, the first NN will not
# be the sample, but a duplicate.
# In that case mask the first duplicate.
dup_gr_nbrs = np.all(sample_mask, axis=1)
sample_mask[:, 0][dup_gr_nbrs] = False
neigh_ind = np.reshape(neigh_ind[sample_mask],
(n_samples, n_neighbors - 1))
if return_distance:
dist = np.reshape(dist[sample_mask],
(n_samples, n_neighbors - 1))
return dist, neigh_ind
return neigh_ind
