def test_cummode():
arr = np.random.RandomState(0).randint(low=0, high=3, size=(5, 7))
m = cummode(arr, axis = 1) # (n_samples, n_events)
assert m.shape == arr.shape
uniques = np.unique(arr)
for j in xrange(arr.shape[1]):
n_elements_of_mode_class = np.sum(arr[:, :j+1] == m[:, j][:, None], axis = 1) # (n_samples, )
for k, u in enumerate(uniques):
n_elements_of_this_class = np.sum(arr[:, :j+1] == u, axis = 1) # (n_samples, )
assert np.all(n_elements_of_mode_class >= n_elements_of_this_class)
def test_cummode():
arr = np.random.RandomState(0).randint(low=0, high=3, size=(5, 7))
m = cummode(arr, axis = 1) # (n_samples, n_events)
assert m.shape == arr.shape
uniques = np.unique(arr)
for j in xrange(arr.shape[1]):
n_elements_of_mode_class = np.sum(arr[:, :j+1] == m[:, j][:, None], axis = 1) # (n_samples, )
for k, u in enumerate(uniques):
n_elements_of_this_class = np.sum(arr[:, :j+1] == u, axis = 1) # (n_samples, )
assert np.all(n_elements_of_mode_class >= n_elements_of_this_class)
def test_cummode_weighted():
arr = np.random.RandomState(0).randint(low=0, high=3, size=(5, 7))
w = np.random.rand(5, 7)
m = cummode(arr, weights=w, axis = 1) # (n_samples, n_events)
assert m.shape == arr.shape
uniques = np.unique(arr)
for j in xrange(arr.shape[1]):
bool_ixs_of_mode_class = arr[:, :j+1] == m[:, j][:, None] # (n_samples, j+1)
weights_of_mode_class = np.sum(w[:, :j+1]*bool_ixs_of_mode_class, axis = 1) # (n_samples, )
for k, u in enumerate(uniques):
bool_ixs_of_this_class = arr[:, :j+1] == u # (n_samples, j+1)
weights_of_this_class = np.sum(w[:, :j+1]*bool_ixs_of_this_class, axis = 1) # (n_samples, )
assert np.all(weights_of_mode_class >= weights_of_this_class)
def test_cummode_weighted():
arr = np.random.RandomState(0).randint(low=0, high=3, size=(5, 7))
w = np.random.rand(5, 7)
m = cummode(arr, weights=w, axis = 1) # (n_samples, n_events)
assert m.shape == arr.shape
uniques = np.unique(arr)
for j in xrange(arr.shape[1]):
bool_ixs_of_mode_class = arr[:, :j+1] == m[:, j][:, None] # (n_samples, j+1)
weights_of_mode_class = np.sum(w[:, :j+1]*bool_ixs_of_mode_class, axis = 1) # (n_samples, )
for k, u in enumerate(uniques):
bool_ixs_of_this_class = arr[:, :j+1] == u # (n_samples, j+1)
weights_of_this_class = np.sum(w[:, :j+1]*bool_ixs_of_this_class, axis = 1) # (n_samples, )
assert np.all(weights_of_mode_class >= weights_of_this_class)