def test_directional_diff():
v = np.r_[1, -1]
v = v / np.linalg.norm(v)
x0 = [2, 3]
directional_diff = np.dot(nd.Gradient(rosen)(x0), v)
assert_allclose(directional_diff, 743.87633380824832)
dd, _info = nd.directionaldiff(rosen, x0, v, full_output=True)
assert_allclose(dd, 743.87633380824832)
def test_issue_39():
"""
Test that checks float/Bicomplex works
"""
fun = nd.Gradient(lambda x: 1.0 / (np.exp(x[0]) + np.cos(x[1]) + 10),
method="multicomplex")
assert_allclose(fun([1.0, 2.0]),
[-0.017961123762187736, 0.0060082083648822])
def test_gradient():
def fun(x):
return np.sum(x**2)
dtrue = [2., 4., 6.]
for method in ['complex', 'central', 'backward', 'forward']:
for order in [2, 4]:
dfun = nd.Gradient(fun, method=method, order=order)
d = dfun([1, 2, 3])
assert_array_almost_equal(d, dtrue)
def test_gradient_fulloutput():
"""Fix issue#52:
Gradient tries to apply squeeze to the output tuple containing both the result
and the full_output object.
"""
res, info = nd.Gradient(lambda x, y: x + y, full_output=True)(1, 3)
assert_allclose(res, 1)
assert info.error_estimate < 1e-13
assert info.final_step == 0.015625
assert info.index == 5