def test_Z3(): """ Test of implementation 2. """ traj = simple_traj10() theta = np.array([500.0, -100.0]) weights = [[0.1, 0.9], [0.9, 0.1]] elts = LearningElementsSecure(traj, theta, weights=weights) elts.computeLogZ()
def test_hess_traj1_2(): """ test_hess_traj1_2 """ traj = simple_traj1() theta = np.array([0.0, -1.0]) choices = [1, 0] elts = LearningElementsSecure(traj, theta, choices) elts.computeHessianLogZ() elts_ref = LearningElementsRef(traj, theta, choices) elts_ref.computeHessianLogZ() h = elts.hess_logZ h_ref = elts_ref.hess_logZ assert(np.abs(h - h_ref).max() < 1e-3), (h, h_ref)
def test_traj_1_2():
""" test_traj_1_2 """
traj = simple_traj1()
theta = np.array([1.0, -1.0])
choices = [1, 0]
elts = LearningElementsSecure(traj, theta, choices)
elts.computeLogZ()
elts_ref = LearningElementsRef(traj, theta, choices)
elts_ref.computeLogZ()
assert(within(elts.Z, elts_ref.Z, 1e-5)), (elts.Z, elts_ref.Z, 1e-5)
assert(within(elts.logZ, elts_ref.logZ, 1e-5)), \
(elts.logZ, elts_ref.logZ, 1e-5)
def test_grad_traj5_2(): """ test_grad_traj5_2 Test of implementation 1 of gradient. """ traj = simple_traj5() theta = np.array([-1.0]) choices = [1, 0, 2] elts = LearningElementsSecure(traj, theta, choices) elts.computeGradientLogZ() elts_ref = LearningElementsRef(traj, theta, choices) elts_ref.computeGradientLogZ() g = elts.grad_logZ g_ref = elts_ref.grad_logZ assert(np.abs(g - g_ref).max() < 1e-3), (g, g_ref)
def test_hess_traj5_2():
""" test_hess_traj5_2 """
traj = simple_traj5()
theta = np.array([-1.0])
choices = [1, 0, 2]
elts = LearningElementsSecure(traj, theta, choices)
elts.computeHessianLogZ()
elts_ref = LearningElementsRef(traj, theta, choices)
elts_ref.computeHessianLogZ()
h = elts.hess_logZ
h_ref = elts_ref.hess_logZ
assert(np.abs(h - h_ref).max() < 1e-3), \
(h, h_ref, elts.hess_Z, elts_ref.hess_Z, \
elts.grad_Z, elts_ref.grad_Z, elts.Z, elts_ref.Z,)
def inner(theta):
""" Returned closure. """
y = 0.0
g = np.zeros_like(theta)
n = len(theta)
h = np.zeros((n, n))
for (traj, weights) in traj_estims:
elts = LearningElements1(traj, theta, weights=weights)
elts.computeValue()
y += elts.logValue
elts.computeGradientValue()
g += elts.grad_logValue
elts.computeHessianValue()
h += elts.hess_logValue
return (y, g, h)
