def test_single_wavelength_gradient_integral(self):
""" Test FOM gradient integral for single wavelength case. """
fom_grad = ModeMatch.fom_gradient_wavelength_integral_impl(
T_fwd_vs_wavelength=np.array([0.2851]),
T_fwd_partial_derivs_vs_wl=np.array([2.0]),
target_T_fwd_vs_wavelength=np.array([1.0]),
wl=np.array([1800e-9]),
norm_p=1)
exact_fom_grad = -1.0 * np.sign(0.2851 - 1.0) * 2.0
self.assertAlmostEqual(fom_grad[0], exact_fom_grad, 15)
def test_fom_gradient_integral_p1(self):
""" Test FOM gradient integral with norm p = 1. """
wl_points = 3
wavelengths = np.linspace(1300e-9, 1800e-9, wl_points)
target_T_fwd = lambda wl: np.linspace(0.0, 1.0, wl.size)
fom_grad = ModeMatch.fom_gradient_wavelength_integral_impl(
T_fwd_vs_wavelength=0.25 * np.ones(wl_points),
T_fwd_partial_derivs_vs_wl=np.ones((1, wl_points)),
target_T_fwd_vs_wavelength=target_T_fwd(wavelengths),
wl=wavelengths,
norm_p=1)
self.assertAlmostEqual(fom_grad[0], 0.5, 15)
def test_fom_gradient_integral_p2(self):
""" Test FOM gradient integral with norm p = 2. """
wl_points = 5000
wavelengths = np.linspace(1.0e-9, 1.0e-8, wl_points)
target_T_fwd = lambda wl: np.exp(-1.0 * (wl - wl.min()) /
(wl.max() - wl.min()))
fom_grad = ModeMatch.fom_gradient_wavelength_integral_impl(
T_fwd_vs_wavelength=0.5 * np.ones(wl_points),
T_fwd_partial_derivs_vs_wl=np.ones((1, wl_points)),
target_T_fwd_vs_wavelength=target_T_fwd(wavelengths),
wl=wavelengths,
norm_p=2)
exact_fom_grad = (np.exp(1.0) - 2.0) / np.sqrt(4.0 * np.exp(1.0) -
np.exp(2.0) - 2.0)
self.assertAlmostEqual(fom_grad[0], exact_fom_grad, 7)
