def test_derivfuncs_dims1():
import ann
x = np.zeros((2,2,2))
y = np.zeros((2,2,2))
MSE = ann.ERROR_MSE
failed = False
try:
ann.get_deriv(MSE, x, y)
except ValueError:
failed = True
assert failed, "Should not work with more than 2 dimensions"
def test_derivfuncs_dims1():
import ann
x = np.zeros((2, 2, 2))
y = np.zeros((2, 2, 2))
MSE = ann.ERROR_MSE
failed = False
try:
ann.get_deriv(MSE, x, y)
except ValueError:
failed = True
assert failed, "Should not work with more than 2 dimensions"
def test_derivfuncs_dims2():
import ann
# First, should fail if arrays differ in dimension
x = np.zeros((2,2))
y = np.zeros((3,3))
MSE = ann.ERROR_MSE
failed = False
try:
ann.get_deriv(MSE, x, y)
except ValueError:
failed = True
assert failed, "Dimensions should not match!"
def test_derivfuncs_dims2():
import ann
# First, should fail if arrays differ in dimension
x = np.zeros((2, 2))
y = np.zeros((3, 3))
MSE = ann.ERROR_MSE
failed = False
try:
ann.get_deriv(MSE, x, y)
except ValueError:
failed = True
assert failed, "Dimensions should not match!"
def test_derivfuncs_data1dlistsminimal():
import ann
rows, cols = 1, 1
x = [0.0 for i in range(rows)]
y = [2.0 for i in range(rows)]
MSE = ann.ERROR_MSE
error = ann.get_deriv(MSE, x, y)
assert len(error.shape) == 1, "Should be one-dimensional result"
assert error.shape[0] == rows, "Count should match row number"
for e in error.ravel():
assert 0.000001 > e - (2 - 0), "Error is incorrect"
def test_derivfuncs_data1d():
import ann
rows, cols = 5, 1
x = np.zeros(rows)
y = np.ones(rows) * 2
MSE = ann.ERROR_MSE
error = ann.get_deriv(MSE, x, y)
assert len(error.shape) == 1, "Should be one-dimensional result"
assert error.shape[0] == rows, "Count should match row number"
for e in error.ravel():
assert 0.000001 > e - (2 - 0), "Error is incorrect"
def test_derivfuncs_data1dlistsminimal():
import ann
rows, cols = 1, 1
x = [0.0 for i in range(rows)]
y = [2.0 for i in range(rows)]
MSE = ann.ERROR_MSE
error = ann.get_deriv(MSE, x, y)
assert len(error.shape) == 1, "Should be one-dimensional result"
assert error.shape[0] == rows, "Count should match row number"
for e in error.ravel():
assert 0.000001 > e - (2 - 0), "Error is incorrect"
def test_derivfuncs_data1d():
import ann
rows, cols = 5, 1
x = np.zeros(rows)
y = np.ones(rows) * 2
MSE = ann.ERROR_MSE
error = ann.get_deriv(MSE, x, y)
assert len(error.shape) == 1, "Should be one-dimensional result"
assert error.shape[0] == rows, "Count should match row number"
for e in error.ravel():
assert 0.000001 > e - (2 - 0), "Error is incorrect"
def test_surv_likelihood():
import ann
dim = (20,2)
targets = np.ones(dim)
censlvl=0.0
for i in range(len(targets)):
if np.random.uniform() < censlvl:
targets[i, 1] = 0
outputs = np.random.normal(1, 10, size=dim)
#timesorting = outputs[:, 0].argsort()
cens = targets[:, 1] < 1
uncens = targets[:, 1] == 1
errors = ann.get_error(ann.ERROR_SURV_LIKELIHOOD, targets, outputs)
derivs = ann.get_deriv(ann.ERROR_SURV_LIKELIHOOD, targets, outputs)
def test_surv_likelihood():
import ann
dim = (20, 2)
targets = np.ones(dim)
censlvl = 0.0
for i in range(len(targets)):
if np.random.uniform() < censlvl:
targets[i, 1] = 0
outputs = np.random.normal(1, 10, size=dim)
#timesorting = outputs[:, 0].argsort()
cens = targets[:, 1] < 1
uncens = targets[:, 1] == 1
errors = ann.get_error(ann.ERROR_SURV_LIKELIHOOD, targets, outputs)
derivs = ann.get_deriv(ann.ERROR_SURV_LIKELIHOOD, targets, outputs)
