def fit(self, data):
"""
Fit the current model with the inputted data.
:param data: Samples to fit model with
:type data: ndarray(double), shape = (n_samples, n_features)
:returns: Trained MiniBatch instance
:rtype: MiniBatch
"""
data = np.asarray(data, dtype=np.double)
if self.verbose:
logger.info("Initializing clusters")
if self.init == 'random':
self.cluster_centers_ = np.random.random(
(self.n_clusters, data.shape[1]), dtype=np.double)
elif self.init == 'kmeans++':
self.cluster_centers_ = np.zeros((self.n_clusters, data.shape[1]),
dtype=np.double)
jobs = min(self.n_jobs, self.n_init)
if jobs > 1:
self.cluster_centers_ = _minibatch.kmeanspp_multi(
data, self.cluster_centers_, self.n_samples, self.n_init,
jobs)
else:
self.cluster_centers_ = _minibatch.kmeanspp(
data, self.cluster_centers_, self.n_samples)
elif isinstance(self.init, np.ndarray):
if not self.init.flags['C_CONTIGUOUS']:
raise TypeError("init ndarray must be C_CONTIGUOUS")
elif self.init.shape != (self.n_clusters, data.shape[1]):
raise TypeError(
"init cluster not of correct shape "
"%r != (%d, %d)" %
(self.init.shape, self.n_clusters, data.shape[1]))
self.cluster_centers_ = self.init
if self.verbose:
logger.info("Running minibatch")
jobs = min(self.n_jobs, self.n_runs)
if jobs > 1:
self.cluster_centers_ = _minibatch.minibatch_multi(
data, self.cluster_centers_, self.n_samples, self.max_iter,
self.n_runs, jobs, self.bic_termination,
self.reassignment_ratio)
else:
self.cluster_centers_ = _minibatch.minibatch(
data, self.cluster_centers_, self.n_samples, self.max_iter,
self.bic_termination, self.reassignment_ratio)
if self.compute_labels:
if self.verbose:
logger.info("Computing labels")
self.labels_ = np.zeros((data.shape[0], ), dtype=np.intc)
self.labels_ = _minibatch.assign_centroids(data,
self.cluster_centers_,
self.labels_,
self.n_jobs)
return self
def fit(self, data):
"""
Fit the current model with the inputted data.
:param data: Samples to fit model with
:type data: ndarray(double), shape = (n_samples, n_features)
:returns: Trained MiniBatch instance
:rtype: MiniBatch
"""
data = np.asarray(data, dtype=np.double)
if self.verbose:
logger.info("Initializing clusters")
if self.init == 'random':
self.cluster_centers_ = np.random.random(
(self.n_clusters, data.shape[1]), dtype=np.double)
elif self.init == 'kmeans++':
self.cluster_centers_ = np.zeros(
(self.n_clusters, data.shape[1]), dtype=np.double)
jobs = min(self.n_jobs, self.n_init)
if jobs > 1:
self.cluster_centers_ = _minibatch.kmeanspp_multi(
data, self.cluster_centers_, self.n_samples,
self.n_init, jobs)
else:
self.cluster_centers_ = _minibatch.kmeanspp(
data, self.cluster_centers_, self.n_samples)
elif isinstance(self.init, np.ndarray):
if not self.init.flags['C_CONTIGUOUS']:
raise TypeError("init ndarray must be C_CONTIGUOUS")
elif self.init.shape != (self.n_clusters, data.shape[1]):
raise TypeError("init cluster not of correct shape "
"%r != (%d, %d)" % (self.init.shape,
self.n_clusters,
data.shape[1]))
self.cluster_centers_ = self.init
if self.verbose:
logger.info("Running minibatch")
jobs = min(self.n_jobs, self.n_runs)
if jobs > 1:
self.cluster_centers_ = _minibatch.minibatch_multi(
data, self.cluster_centers_, self.n_samples, self.max_iter,
self.n_runs, jobs, self.bic_termination,
self.reassignment_ratio)
else:
self.cluster_centers_ = _minibatch.minibatch(
data, self.cluster_centers_, self.n_samples, self.max_iter,
self.bic_termination, self.reassignment_ratio)
if self.compute_labels:
if self.verbose:
logger.info("Computing labels")
self.labels_ = np.zeros((data.shape[0], ), dtype=np.intc)
self.labels_ = _minibatch.assign_centroids(
data, self.cluster_centers_, self.labels_, self.n_jobs)
return self
def predict(self, data):
"""
Labels the data given the fitted mode.
:param data: Samples to classify
:type data: ndarray(double), shape = (n_samples, n_features)
:returns: Index into MiniBatch.cluster_centers_ for each datapoint in
data
:rtype: ndarray(intc), shape = (n_samples,)
"""
assert self.cluster_centers_ is not None, "Model not yet fitted"
labels = np.zeros((data.shape[0], ), dtype=np.intc)
labels = _minibatch.assign_centroids(data, self.cluster_centers_, labels, self.n_jobs)
return labels
def predict(self, data):
"""
Labels the data given the fitted mode.
:param data: Samples to classify
:type data: ndarray(double), shape = (n_samples, n_features)
:returns: Index into MiniBatch.cluster_centers_ for each datapoint in
data
:rtype: ndarray(intc), shape = (n_samples,)
"""
assert self.cluster_centers_ is not None, "Model not yet fitted"
labels = np.zeros((data.shape[0], ), dtype=np.intc)
labels = _minibatch.assign_centroids(data, self.cluster_centers_,
labels, self.n_jobs)
return labels
