def test_forward(self):
inputs = torch.Tensor([-6, -5, -3, -1, 0, 1, 3, 5])
output = gpytorch.log_normal_cdf(Variable(inputs)).data
# Answers should be reasonable for small values
self.assertLess(math.fabs(output[0] + 20.7368), 1e-4)
self.assertLess(math.fabs(output[1] + 15), 0.1)
self.assertLess(math.fabs(output[2] + 6.6), 0.01)
self.assertLess(math.fabs(output[3] + 1.841), 0.001)
# Should be very accurate for positive values
self.assertLess(math.fabs(output[4] + 0.693147), 1e-4)
self.assertLess(math.fabs(output[5] + 0.1727), 1e-4)
self.assertLess(math.fabs(output[6] + 0.00135081), 1e-4)
self.assertLess(math.fabs(output[7] + 2.86652e-7), 1e-4)
def test_backward(self):
inputs = nn.Parameter(torch.Tensor([-6, -5, -3, -1, 0, 1, 3, 5]))
output = gpytorch.log_normal_cdf(inputs)
output.backward(torch.ones(8))
gradient = inputs.grad.data
expected_gradient = torch.Tensor([6.1585, 5.1865, 3.2831, 1.5251, 0.7979, 0.2876, 0.0044, 0.0000])
# Should be reasonable for small values
for d in torch.abs(gradient[:3] - expected_gradient[:3]):
self.assertLess(d, 5e-1)
# Should be very accurate for larger ones
for d in torch.abs(gradient[3:] - expected_gradient[3:]):
self.assertLess(d, 5e-4)