def fit(self):
if self.is_fit:
return
if self.mode == 'transform':
# DONT run the fit, just load a prefitted model from disk
with tables.open_file(self.load_from) as f:
self.kcenters = KCenters.from_pytables(f.root.KCenters)
else:
dataset = []
for data, fn in self._yield_input(with_filenames=True):
self.breakpoints.append(sum(self.breakpoints) + len(data))
self.traj_filenames.append(fn)
dataset.append(data)
dataset = np.concatenate(dataset)
self.log.info('** Starting fitting KCenters...')
self.kcenters.fit(dataset)
self.log.info('** Finished fitting KCenters')
self.is_fit = True
def test_kcenters_1():
data = np.random.randn(100,4)
k = KCenters(10).fit(data)
np.testing.assert_array_equal(k.labels_, k.predict(data))
np.testing.assert_array_equal(k.labels_[k.center_indices_], np.arange(10))
