def __init__(self, lens_data, tracer, padded_tracer):
""" An abstract lens profile fitter, which generates the image-plane image of all galaxies (with light \
profiles) in the tracer and blurs it with the lens data's PSF.
If a padded tracer is supplied, the blurred profile image's can be generated over the entire image and thus \
without the mask.
Parameters
-----------
lens_data : lens_data.LensData
The lens-image that is fitted.
tracer : ray_tracing.AbstractTracerNonStack
The tracer, which describes the ray-tracing and strong lens configuration.
padded_tracer : ray_tracing.Tracer or None
A tracer with an identical strong lens configuration to the tracer above, but using the lens data's \
padded grid_stack such that unmasked model-images can be computed.
"""
super(AbstractLensProfileFit,
self).__init__(tracer=tracer,
padded_tracer=padded_tracer,
psf=lens_data.psf,
map_to_scaled_array=lens_data.map_to_scaled_array)
self.convolver_image = lens_data.convolver_image
blurred_profile_image_1d = util.blurred_image_1d_from_1d_unblurred_and_blurring_images(
unblurred_image_1d=tracer.image_plane_image_1d,
blurring_image_1d=tracer.image_plane_blurring_image_1d,
convolver=self.convolver_image)
self.blurred_profile_image = self.map_to_scaled_array(
array_1d=blurred_profile_image_1d)
def blurred_images_1d_of_images_from_1d_unblurred_and_bluring_images(
unblurred_images_1d, blurring_images_1d, convolvers):
"""For a list of 1D masked images and 1D blurring images (the regions outside the masks whose light blurs \
into the masks after PSF convolution), use both to compute the blurred image within the masks via PSF convolution.
The convolution uses each image's convolver (*See ccd.convolution*).
Parameters
----------
unblurred_images_1d : [ndarray]
List of the 1D masked images which are blurred.
blurring_images_1d : [ndarray]
List of the 1D masked blurring images which are used for blurring.
convolvers : [ccd.convolution.ConvolverImage]
List of the image-convolvers which perform the convolutions in 1D.
"""
blurred_images = []
for image_index in range(len(unblurred_images_1d)):
blurred_profile_image = lens_fit_util.blurred_image_1d_from_1d_unblurred_and_blurring_images(
unblurred_image_1d=unblurred_images_1d[image_index],
blurring_image_1d=blurring_images_1d[image_index],
convolver=convolvers[image_index])
blurred_images.append(blurred_profile_image)
return blurred_images
def test__tracer_and_tracer_sensitive_are_identical__added__likelihood_is_noise_term(
self, lens_data_blur):
g0 = g.Galaxy(mass_profile=mp.SphericalIsothermal(einstein_radius=1.0))
g1 = g.Galaxy(light_profile=lp.EllipticalSersic(intensity=2.0))
tracer = ray_tracing.TracerImageSourcePlanes(
lens_galaxies=[g0],
source_galaxies=[g1],
image_plane_grid_stack=lens_data_blur.grid_stack)
fit = sensitivity_fit.SensitivityProfileFit(lens_data=lens_data_blur,
tracer_normal=tracer,
tracer_sensitive=tracer)
assert (fit.fit_normal.image == lens_data_blur.image).all()
assert (fit.fit_normal.noise_map == lens_data_blur.noise_map).all()
model_image_1d = util.blurred_image_1d_from_1d_unblurred_and_blurring_images(
unblurred_image_1d=tracer.image_plane_image_1d,
blurring_image_1d=tracer.image_plane_blurring_image_1d,
convolver=lens_data_blur.convolver_image)
model_image = lens_data_blur.map_to_scaled_array(
array_1d=model_image_1d)
assert (fit.fit_normal.model_image == model_image).all()
residual_map = fit_util.residual_map_from_data_mask_and_model_data(
data=lens_data_blur.image,
mask=lens_data_blur.mask,
model_data=model_image)
assert (fit.fit_normal.residual_map == residual_map).all()
chi_squared_map = fit_util.chi_squared_map_from_residual_map_noise_map_and_mask(
residual_map=residual_map,
mask=lens_data_blur.mask,
noise_map=lens_data_blur.noise_map)
assert (fit.fit_normal.chi_squared_map == chi_squared_map).all()
assert (fit.fit_sensitive.image == lens_data_blur.image).all()
assert (fit.fit_sensitive.noise_map == lens_data_blur.noise_map).all()
assert (fit.fit_sensitive.model_image == model_image).all()
assert (fit.fit_sensitive.residual_map == residual_map).all()
assert (fit.fit_sensitive.chi_squared_map == chi_squared_map).all()
chi_squared = fit_util.chi_squared_from_chi_squared_map_and_mask(
chi_squared_map=chi_squared_map, mask=lens_data_blur.mask)
noise_normalization = fit_util.noise_normalization_from_noise_map_and_mask(
mask=lens_data_blur.mask, noise_map=lens_data_blur.noise_map)
assert fit.fit_normal.likelihood == -0.5 * (chi_squared +
noise_normalization)
assert fit.fit_sensitive.likelihood == -0.5 * (chi_squared +
noise_normalization)
assert fit.figure_of_merit == 0.0
def __init__(self, lens_data, noise_map_1d, tracer, padded_tracer):
""" An abstract lens profile and inversion fitter, which first generates and subtracts the image-plane \
image of all galaxies (with light profiles) in the tracer, blurs it with the PSF and fits the residual image \
with an inversion using the mapper(s) and regularization(s) in the galaxy's of the tracer.
This inversion use's the lens-image, its PSF and an input noise-map.
If a padded tracer is supplied, the blurred profile image's can be generated over the entire image and thus \
without the mask.
Parameters
-----------
lens_data : lens_data.LensData
The lens-image that is fitted.
noise_map_1d : ndarray
The 1D noise_map map that is fitted, which is an input variable so a hyper-noise_map map can be used (see \
*AbstractHyperFitter*).
tracer : ray_tracing.Tracer
The tracer, which describes the ray-tracing and strong lens configuration.
padded_tracer : ray_tracing.AbstractTracerNonStack or None
A tracer with an identical strong lens configuration to the tracer above, but using the lens data's \
padded grid_stack such that unmasked model-images can be computed.
"""
super(AbstractLensProfileInversionFit,
self).__init__(tracer=tracer,
padded_tracer=padded_tracer,
psf=lens_data.psf,
map_to_scaled_array=lens_data.map_to_scaled_array)
self.psf = lens_data.psf
self.convolver_image = lens_data.convolver_image
blurred_profile_image_1d = util.blurred_image_1d_from_1d_unblurred_and_blurring_images(
unblurred_image_1d=tracer.image_plane_image_1d,
blurring_image_1d=tracer.image_plane_blurring_image_1d,
convolver=lens_data.convolver_image)
self.blurred_profile_image = self.map_to_scaled_array(
array_1d=blurred_profile_image_1d)
profile_subtracted_image_1d = lens_data.image_1d - blurred_profile_image_1d
self.profile_subtracted_image = lens_data.image - self.blurred_profile_image
self.inversion = inversions.inversion_from_image_mapper_and_regularization(
image_1d=profile_subtracted_image_1d,
noise_map_1d=noise_map_1d,
convolver=lens_data.convolver_mapping_matrix,
mapper=tracer.mappers_of_planes[-1],
regularization=tracer.regularizations_of_planes[-1])
def test___manual_image_and_psf(self, lens_data_stack_manual):
g0 = g.Galaxy(light_profile=lp.EllipticalSersic(intensity=1.0))
g1 = g.Galaxy(mass_profile=mp.SphericalIsothermal(
einstein_radius=1.0))
tracer = ray_tracing_stack.TracerImageSourcePlanesStack(
lens_galaxies=[g0, g1],
source_galaxies=[g0],
image_plane_grid_stacks=lens_data_stack_manual.grid_stacks)
padded_tracer = ray_tracing_stack.TracerImageSourcePlanesStack(
lens_galaxies=[g0, g1],
source_galaxies=[g0],
image_plane_grid_stacks=lens_data_stack_manual.
padded_grid_stacks)
fit = lens_fit_stack.fit_lens_image_stack_with_tracer(
lens_data_stack=lens_data_stack_manual,
tracer=tracer,
padded_tracer=padded_tracer)
assert lens_data_stack_manual.noise_maps[0] == pytest.approx(
fit.noise_maps[0], 1e-4)
assert lens_data_stack_manual.noise_maps[1] == pytest.approx(
fit.noise_maps[1], 1e-4)
model_image_1d_0 = lens_fit_util.blurred_image_1d_from_1d_unblurred_and_blurring_images(
unblurred_image_1d=tracer.image_plane_images_1d[0],
blurring_image_1d=tracer.image_plane_blurring_images_1d[0],
convolver=lens_data_stack_manual.convolvers_image[0])
model_image_1d_1 = lens_fit_util.blurred_image_1d_from_1d_unblurred_and_blurring_images(
unblurred_image_1d=tracer.image_plane_images_1d[1],
blurring_image_1d=tracer.image_plane_blurring_images_1d[1],
convolver=lens_data_stack_manual.convolvers_image[1])
model_image_0 = lens_data_stack_manual.map_to_scaled_arrays[0](
array_1d=model_image_1d_0)
model_image_1 = lens_data_stack_manual.map_to_scaled_arrays[1](
array_1d=model_image_1d_1)
assert model_image_0 == pytest.approx(fit.model_images[0], 1e-4)
assert model_image_1 == pytest.approx(fit.model_images[1], 1e-4)
residual_map_0 = fit_util.residual_map_from_data_mask_and_model_data(
data=lens_data_stack_manual.images[0],
mask=lens_data_stack_manual.masks[0],
model_data=model_image_0)
residual_map_1 = fit_util.residual_map_from_data_mask_and_model_data(
data=lens_data_stack_manual.images[1],
mask=lens_data_stack_manual.masks[1],
model_data=model_image_1)
assert residual_map_0 == pytest.approx(fit.residual_maps[0], 1e-4)
assert residual_map_1 == pytest.approx(fit.residual_maps[1], 1e-4)
chi_squared_map_0 = fit_util.chi_squared_map_from_residual_map_noise_map_and_mask(
residual_map=residual_map_0,
mask=lens_data_stack_manual.masks[0],
noise_map=lens_data_stack_manual.noise_maps[0])
chi_squared_map_1 = fit_util.chi_squared_map_from_residual_map_noise_map_and_mask(
residual_map=residual_map_1,
mask=lens_data_stack_manual.masks[1],
noise_map=lens_data_stack_manual.noise_maps[1])
assert chi_squared_map_0 == pytest.approx(fit.chi_squared_maps[0],
1e-4)
assert chi_squared_map_1 == pytest.approx(fit.chi_squared_maps[1],
1e-4)
chi_squared_0 = fit_util.chi_squared_from_chi_squared_map_and_mask(
chi_squared_map=chi_squared_map_0,
mask=lens_data_stack_manual.masks[0])
noise_normalization_0 = fit_util.noise_normalization_from_noise_map_and_mask(
noise_map=lens_data_stack_manual.noise_maps[0],
mask=lens_data_stack_manual.masks[0])
likelihood_0 = fit_util.likelihood_from_chi_squared_and_noise_normalization(
chi_squared=chi_squared_0,
noise_normalization=noise_normalization_0)
chi_squared_1 = fit_util.chi_squared_from_chi_squared_map_and_mask(
chi_squared_map=chi_squared_map_1,
mask=lens_data_stack_manual.masks[1])
noise_normalization_1 = fit_util.noise_normalization_from_noise_map_and_mask(
noise_map=lens_data_stack_manual.noise_maps[1],
mask=lens_data_stack_manual.masks[1])
likelihood_1 = fit_util.likelihood_from_chi_squared_and_noise_normalization(
chi_squared=chi_squared_1,
noise_normalization=noise_normalization_1)
assert fit.chi_squareds == [chi_squared_0, chi_squared_1]
assert fit.noise_normalizations == [
noise_normalization_0, noise_normalization_1
]
assert fit.likelihoods == [likelihood_0, likelihood_1]
assert likelihood_0 + likelihood_1 == pytest.approx(
fit.likelihood, 1e-4)
assert likelihood_0 + likelihood_1 == fit.figure_of_merit
blurred_image_of_planes_0 = lens_fit_util.blurred_image_of_planes_from_1d_images_and_convolver(
total_planes=tracer.total_planes,
image_plane_image_1d_of_planes=tracer.
image_plane_images_1d_of_planes[0],
image_plane_blurring_image_1d_of_planes=tracer.
image_plane_blurring_images_1d_of_planes[0],
convolver=lens_data_stack_manual.convolvers_image[0],
map_to_scaled_array=lens_data_stack_manual.
map_to_scaled_arrays[0])
blurred_image_of_planes_1 = lens_fit_util.blurred_image_of_planes_from_1d_images_and_convolver(
total_planes=tracer.total_planes,
image_plane_image_1d_of_planes=tracer.
image_plane_images_1d_of_planes[1],
image_plane_blurring_image_1d_of_planes=tracer.
image_plane_blurring_images_1d_of_planes[1],
convolver=lens_data_stack_manual.convolvers_image[1],
map_to_scaled_array=lens_data_stack_manual.
map_to_scaled_arrays[1])
assert (blurred_image_of_planes_0[0] ==
fit.model_images_of_planes[0][0]).all()
assert (blurred_image_of_planes_0[1] ==
fit.model_images_of_planes[0][1]).all()
assert (blurred_image_of_planes_1[0] ==
fit.model_images_of_planes[1][0]).all()
assert (blurred_image_of_planes_1[1] ==
fit.model_images_of_planes[1][1]).all()
unmasked_blurred_image_0 = \
lens_fit_util.unmasked_blurred_image_from_padded_grid_stack_psf_and_unmasked_image(
padded_grid_stack=lens_data_stack_manual.padded_grid_stacks[0], psf=lens_data_stack_manual.psfs[0],
unmasked_image_1d=padded_tracer.image_plane_images_1d[0])
unmasked_blurred_image_1 = \
lens_fit_util.unmasked_blurred_image_from_padded_grid_stack_psf_and_unmasked_image(
padded_grid_stack=lens_data_stack_manual.padded_grid_stacks[1], psf=lens_data_stack_manual.psfs[1],
unmasked_image_1d=padded_tracer.image_plane_images_1d[1])
assert (unmasked_blurred_image_0 == fit.unmasked_model_images[0]
).all()
assert (unmasked_blurred_image_1 == fit.unmasked_model_images[1]
).all()
unmasked_blurred_image_of_galaxies_i0 = \
lens_fit_util.unmasked_blurred_image_of_galaxies_from_psf_and_unmasked_1d_galaxy_images(
galaxies=padded_tracer.image_plane.galaxies,
image_plane_image_1d_of_galaxies=padded_tracer.image_plane.image_plane_images_1d_of_galaxies[0],
padded_grid_stack=lens_data_stack_manual.padded_grid_stacks[0], psf=lens_data_stack_manual.psfs[0])
unmasked_blurred_image_of_galaxies_s0 = \
lens_fit_util.unmasked_blurred_image_of_galaxies_from_psf_and_unmasked_1d_galaxy_images(
galaxies=padded_tracer.source_plane.galaxies,
image_plane_image_1d_of_galaxies=padded_tracer.source_plane.image_plane_images_1d_of_galaxies[0],
padded_grid_stack=lens_data_stack_manual.padded_grid_stacks[0], psf=lens_data_stack_manual.psfs[0])
unmasked_blurred_image_of_galaxies_i1 = \
lens_fit_util.unmasked_blurred_image_of_galaxies_from_psf_and_unmasked_1d_galaxy_images(
galaxies=padded_tracer.image_plane.galaxies,
image_plane_image_1d_of_galaxies=padded_tracer.image_plane.image_plane_images_1d_of_galaxies[0],
padded_grid_stack=lens_data_stack_manual.padded_grid_stacks[1], psf=lens_data_stack_manual.psfs[1])
unmasked_blurred_image_of_galaxies_s1 = \
lens_fit_util.unmasked_blurred_image_of_galaxies_from_psf_and_unmasked_1d_galaxy_images(
galaxies=padded_tracer.source_plane.galaxies,
image_plane_image_1d_of_galaxies=padded_tracer.source_plane.image_plane_images_1d_of_galaxies[0],
padded_grid_stack=lens_data_stack_manual.padded_grid_stacks[1], psf=lens_data_stack_manual.psfs[1])
assert (
unmasked_blurred_image_of_galaxies_i0[0] == fit.
unmasked_model_images_of_planes_and_galaxies[0][0][0]).all()
assert (
unmasked_blurred_image_of_galaxies_s0[0] == fit.
unmasked_model_images_of_planes_and_galaxies[0][1][0]).all()
assert (
unmasked_blurred_image_of_galaxies_i1[0] == fit.
unmasked_model_images_of_planes_and_galaxies[1][0][0]).all()
assert (
unmasked_blurred_image_of_galaxies_s1[0] == fit.
unmasked_model_images_of_planes_and_galaxies[1][1][0]).all()
start_overall = time.time()
start = time.time()
for i in range(repeats):
tracer = ray_tracing.TracerImageSourcePlanes(
lens_galaxies=[lens_galaxy],
source_galaxies=[source_galaxy],
image_plane_grid_stack=lens_data.grid_stack)
diff = time.time() - start
print("Time to Setup Tracer = {}".format(diff / repeats))
start = time.time()
for i in range(repeats):
blurred_profile_image_1d = lens_fit_util.blurred_image_1d_from_1d_unblurred_and_blurring_images(
unblurred_image_1d=tracer.image_plane_image_1d,
blurring_image_1d=tracer.image_plane_blurring_image_1d,
convolver=lens_data.convolver_image)
blurred_profile_image = lens_data.map_to_scaled_array(
array_1d=blurred_profile_image_1d)
diff = time.time() - start
print("Time to perform PSF convolution = {}".format(diff / repeats))
start = time.time()
for i in range(repeats):
lens_fit.LensDataFit(image=lens_data.image_1d,
noise_map=lens_data.noise_map_1d,
mask=lens_data.mask_1d,
model_image=blurred_profile_image_1d)
diff = time.time() - start
print("Time to perform fit (1D) = {}".format(diff / repeats))
"\n")
print("Number of points = " + str(lens_data.grid.shape[0]) + "\n")
start_overall = time.time()
start = time.time()
for i in range(repeats):
tracer = al.Tracer.from_galaxies(galaxies=[lens_galaxy, source_galaxy])
diff = time.time() - start
print("Time to Setup Tracer = {}".format(diff / repeats))
start = time.time()
for i in range(repeats):
blurred_profile_image_1d = lens_fit_util.blurred_image_1d_from_1d_unblurred_and_blurring_images(
unblurred_image_1d=tracer.profile_image_from_grid,
blurring_image_1d=tracer.profile_blurring_image,
convolver=lens_data.convolver,
)
blurred_profile_image = lens_data.grid.mapping.scaled_array_2d_from_array_1d(
array_1d=blurred_profile_image_1d)
diff = time.time() - start
print("Time to perform PSF convolution = {}".format(diff / repeats))
start = time.time()
for i in range(repeats):
al.LensDataFit(
image_1d=lens_data.image_1d,
noise_map_1d=lens_data.noise_map_1d,
mask_1d=lens_data.mask_1d,
model_image_1d=blurred_profile_image_1d,
)