def test_longer(self):
ass = Assembly()
ass.coalesce(np.array([[0., 0.], [2., 2.]]))
labels = ass.coalesce(np.array([[1.9, 1.9], [3., 3.], [0.1, 0.1]]))
self.assertTrue(np.array_equal(labels, [1, 2, 0]))
self.assertTrue(
np.array_equal(ass.points, [[0.1, 0.1], [1.9, 1.9], [3., 3.]]))
def test_coalesce(self):
ass = Assembly()
ass.coalesce(np.array([[0., 0.], [2., 2.]]))
ass.coalesce(np.array([[1.9, 1.9], [3, 3], [0.1, 0.1]]))
labels = ass.coalesce(np.array([[2.9, 2.9], [2.1, 2.1]]))
self.assertTrue(np.array_equal(labels, [2, 1]))
self.assertTrue(
np.array_equal(ass.points, [[0.1, 0.1], [2.1, 2.1], [2.9, 2.9]]))
class OC:
"""Wraps a mini batch MCMC algorithm"""
def __init__(self, **kwargs):
"""
Build a new OC instance
"""
frame_size = kwargs.get('frame_size', 100)
self.frame = deque([], frame_size)
del kwargs['frame_size']
self.algo = Batch(MCMC, **kwargs)
self.ass = Assembly()
def push_predict(self, points, columns):
"""
push and predict data from the dataflow
:param points: a chunk of data
:param columns: the column names
:return: the result that can be send to the dataviz server
"""
arr = self._extend(points)
self.algo.push(arr)
centroids, labels = self.algo.predict(arr)
return self._make_result(centroids, labels, columns)
def _extend(self, points):
"""
keep the given chunk in the fixed size frame. Old data can be reused
if the chunk is smaller than the frame.
:param points: a chunk of data
:return: the content of the frame as a ndarray
"""
self.frame.extend(points)
return np.array(self.frame)
def _make_result(self, centroids, labels, columns):
"""
Build a result ready for the dataviz server. The result contains
all known centers, even if empty for the current centroids
:param centroids: the current centroids
:param labels: the labels for the data in the frame
:param columns: the column names
:return: the result ready fo the dataviz server
"""
indices = self.ass.coalesce(centroids)
counts = np.zeros(len(self.ass.points))
counts[indices] = np.bincount(labels)
return {
'centers': self.ass.points.tolist(),
'counts': counts.tolist(),
'columns': columns
}
def run(self):
"""Run the underlying algorithm"""
return self.algo.run()
def test_init(self):
ass = Assembly()
labels = ass.coalesce(np.array([[0., 0.], [2., 2.]]))
self.assertTrue(np.array_equal(labels, [0, 1]))
def test_shorter(self):
ass = Assembly()
ass.coalesce(np.array([[0., 0.], [2., 2.]]))
labels = ass.coalesce(np.array([[1.9, 1.9]]))
self.assertTrue(np.array_equal(labels, [1]))
self.assertTrue(np.array_equal(ass.points, [[0., 0.], [1.9, 1.9]]))
