def test_vstack_with_single_vector(self):
expected = SparseMatrix.from_list([[1, 0, 2, 0, 3, 4, 0, 5]])
result = SparseMatrix.vstack(
[SparseVector.from_list([1, 0, 2, 0, 3, 4, 0, 5])])
self.assertEqual(result.shape, (1, 8))
self.assertEqual(result, expected)
def test_vstack_with_multiple_vectors(self):
arrays = [np.random.randint(0, 10, 30) for _ in range(20)]
vectors = [SparseVector.from_list(arrays[i]) for i in range(20)]
expected = SparseMatrix.from_list(arrays)
result = SparseMatrix.vstack(vectors)
self.assertEqual(result, expected)
def select(self, classz):
"""
Computes the sub-matrix that represents all training examples whose
classification is the specified class.
:param classz: The class to select.
:return: A sparse matrix of data specific to a class.
"""
indices = np.where(self.classes == classz)[0]
return SparseMatrix.vstack([self.counts.get_row(i) for i in indices])
def shuffle(self):
"""
Shuffles this training database, re-arranging the order of both the
classes and data by row only.
"""
rows = [row for row in self.counts.get_rows()]
state = np.random.get_state()
for item in [self.classes, rows]:
np.random.set_state(state)
np.random.shuffle(item)
self.counts = SparseMatrix.vstack(rows, dtype=self.counts.data.dtype)
