def __init__(self, *args, lmbda_1=0., lmbda_2=0.):
super().__init__(*args)
self.y = asarray2d(self.y)
if lmbda_1 <= 0:
lmbda_1 = estimate_entropy(self.y) / LAMBDA_1_ADJUSTMENT
if lmbda_2 <= 0:
lmbda_2 = estimate_entropy(self.y) / LAMBDA_2_ADJUSTMENT
self.lmbda_1 = lmbda_1
self.lmbda_2 = lmbda_2
def test_entropy_multiple_disc():
same_val_arr_zero = np.zeros((50, 1))
same_val_arr_ones = np.ones((50, 1))
# The 0.5 forces float => classified as continuous
cont_val_arange = asarray2d(np.arange(50) + 0.5)
all_disc_arr = np.concatenate((same_val_arr_ones, same_val_arr_zero),
axis=1)
mixed_val_arr = np.concatenate((all_disc_arr, cont_val_arange), axis=1)
all_disc_h = estimate_entropy(all_disc_arr)
mixed_h = estimate_entropy(mixed_val_arr)
assert mixed_h > all_disc_h, \
'Expected adding continuous column increases entropy'
def __init__(self,
*args,
lmbda_1: float = 0.0,
lmbda_2: float = 0.0,
lambda_1_adjustment: float = LAMBDA_1_ADJUSTMENT,
lambda_2_adjustment: float = LAMBDA_2_ADJUSTMENT):
super().__init__(*args)
self.y_val = asarray2d(self.y_val)
if lmbda_1 <= 0:
lmbda_1 = estimate_entropy(self.y_val) / lambda_1_adjustment
if lmbda_2 <= 0:
lmbda_2 = estimate_entropy(self.y_val) / lambda_2_adjustment
self.lmbda_1 = lmbda_1
self.lmbda_2 = lmbda_2
def test_mi_informative():
x = np.reshape(np.arange(1, 101), (-1, 1))
y = np.reshape(np.arange(1, 101), (-1, 1))
mi = estimate_mutual_information(x, y)
h_y = estimate_entropy(y)
assert mi > h_y / 4, \
'exact copy columns should have high information'
def test_mi_uninformative():
x = np.reshape(np.arange(1, 101), (-1, 1))
y = np.ones((100, 1))
mi = estimate_mutual_information(x, y)
h_z = estimate_entropy(x)
assert h_z / 4 > mi, \
'uninformative column should have no information'
def test_mi_informative(self):
x = np.reshape(np.arange(1, 101), (-1, 1))
y = np.reshape(np.arange(1, 101), (-1, 1))
mi = estimate_mutual_information(x, y)
h_y = estimate_entropy(y)
self.assertGreater(mi, h_y / 4,
'exact copy columns should have high information')
def test_mi_uninformative(self):
x = np.reshape(np.arange(1, 101), (-1, 1))
y = np.ones((100, 1))
mi = estimate_mutual_information(x, y)
h_z = estimate_entropy(x)
self.assertGreater(h_z / 4, mi,
'uninformative column should have no information')
def test_cmi_redundant_info():
x = np.reshape(np.arange(1, 101), (-1, 1))
y = np.reshape(np.arange(1, 101), (-1, 1))
exact_z = np.reshape(np.arange(1, 101), (-1, 1))
h_y = estimate_entropy(y)
cmi = estimate_conditional_information(x, y, exact_z)
assert h_y / 4 > cmi, \
'redundant copies should have little information'
def H(a): # noqa
return estimate_entropy(asarray2d(a))