def test__samples_from_model(self):
lbfgs = af.LBFGS()
lbfgs.paths = af.DirectoryPaths(path_prefix=path.join("non_linear", "LBFGS"))
lbfgs.paths._identifier = "tag"
model = af.ModelMapper(mock_class=af.m.MockClassx3)
model.mock_class.one = af.LogUniformPrior(lower_limit=1e-8, upper_limit=100.0)
model.mock_class.two = af.LogUniformPrior(lower_limit=1e-8, upper_limit=100.0)
model.mock_class.three = af.LogUniformPrior(lower_limit=1e-8, upper_limit=100.0)
samples = lbfgs.samples_from(model=model)
assert isinstance(samples.parameter_lists, list)
assert isinstance(samples.parameter_lists[0], list)
assert isinstance(samples.log_likelihood_list, list)
assert isinstance(samples.log_prior_list, list)
assert isinstance(samples.log_posterior_list, list)
assert samples.parameter_lists[0] == pytest.approx(
[50.005469, 25.143677, 10.06950], 1.0e-4
)
assert samples.log_likelihood_list[0] == pytest.approx(-45.134121, 1.0e-4)
assert samples.log_posterior_list[0] == pytest.approx(-44.97504284, 1.0e-4)
assert samples.weight_list[0] == 1.0
assert len(samples.parameter_lists) == 1
assert len(samples.log_likelihood_list) == 1
def test__samples_from_model(self):
pyswarms = af.PySwarmsGlobal()
pyswarms.paths = af.DirectoryPaths(
path_prefix=path.join("non_linear", "pyswarms"))
pyswarms.paths._identifier = "tag"
model = af.ModelMapper(mock_class=af.m.MockClassx3)
model.mock_class.one = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=100.0)
model.mock_class.two = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=100.0)
model.mock_class.three = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=100.0)
# model.mock_class.four = af.LogUniformPrior(lower_limit=1e-8, upper_limit=100.0)
samples = pyswarms.samples_from(model=model)
assert isinstance(samples.parameter_lists, list)
assert isinstance(samples.parameter_lists[0], list)
assert isinstance(samples.log_likelihood_list, list)
assert isinstance(samples.log_prior_list, list)
assert isinstance(samples.log_posterior_list, list)
assert samples.parameter_lists[0] == pytest.approx(
[50.1254, 1.04626, 10.09456], 1.0e-4)
assert samples.log_likelihood_list[0] == pytest.approx(
-5071.80777, 1.0e-4)
assert samples.log_posterior_list[0] == pytest.approx(
-5070.73298, 1.0e-4)
assert samples.weight_list[0] == 1.0
assert len(samples.parameter_lists) == 500
assert len(samples.log_likelihood_list) == 500
def test__log_prior_from_value(self):
gaussian_simple = af.LogUniformPrior(lower_limit=1e-8, upper_limit=1.0)
log_prior = gaussian_simple.log_prior_from_value(value=1.0)
assert log_prior == 1.0
log_prior = gaussian_simple.log_prior_from_value(value=2.0)
assert log_prior == 0.5
log_prior = gaussian_simple.log_prior_from_value(value=4.0)
assert log_prior == 0.25
gaussian_simple = af.LogUniformPrior(lower_limit=50.0,
upper_limit=100.0)
log_prior = gaussian_simple.log_prior_from_value(value=1.0)
assert log_prior == 1.0
log_prior = gaussian_simple.log_prior_from_value(value=2.0)
assert log_prior == 0.5
log_prior = gaussian_simple.log_prior_from_value(value=4.0)
assert log_prior == 0.25
def test__samples_from_model(self):
drawer = af.Drawer()
drawer.paths = af.DirectoryPaths(
path_prefix=path.join("non_linear", "drawer"))
drawer.paths._identifier = "tag"
model = af.ModelMapper(mock_class=af.m.MockClassx3)
model.mock_class.one = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=100.0)
model.mock_class.two = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=100.0)
model.mock_class.three = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=100.0)
samples = drawer.samples_from(model=model)
assert isinstance(samples.parameter_lists, list)
assert isinstance(samples.parameter_lists[0], list)
assert isinstance(samples.log_likelihood_list, list)
assert isinstance(samples.log_prior_list, list)
assert isinstance(samples.log_posterior_list, list)
assert samples.parameter_lists[0] == pytest.approx(
[49.507679, 49.177471, 14.76753], 1.0e-4)
assert samples.log_likelihood_list[0] == pytest.approx(
-2763.925766, 1.0e-4)
assert samples.log_posterior_list[0] == pytest.approx(
-2763.817517, 1.0e-4)
assert samples.weight_list[0] == 1.0
assert len(samples.parameter_lists) == 3
assert len(samples.log_likelihood_list) == 3
def test__lower_limit_zero_or_below_raises_error(self):
with pytest.raises(exc.PriorException):
af.LogUniformPrior(lower_limit=-1.0, upper_limit=1.0)
with pytest.raises(exc.PriorException):
af.LogUniformPrior(lower_limit=0.0, upper_limit=1.0)
def test__samples_from_model(self):
emcee = af.Emcee()
emcee.paths = af.DirectoryPaths(
path_prefix=path.join("non_linear", "emcee"))
emcee.paths._identifier = "tag"
model = af.ModelMapper(mock_class=af.m.MockClassx4)
model.mock_class.two = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=10.0)
samples = emcee.samples_from(model=model)
assert isinstance(samples.parameter_lists, list)
assert isinstance(samples.parameter_lists[0], list)
assert isinstance(samples.log_likelihood_list, list)
assert isinstance(samples.log_prior_list, list)
assert isinstance(samples.log_posterior_list, list)
assert isinstance(samples.weight_list, list)
assert samples.parameter_lists[0] == pytest.approx(
[0.173670, 0.162607, 3095.28, 0.62104], 1.0e-4)
assert samples.log_likelihood_list[0] == pytest.approx(
-17257775239.32677, 1.0e-4)
assert samples.log_prior_list[0] == pytest.approx(
1.6102016075510708, 1.0e-4)
assert samples.weight_list[0] == pytest.approx(1.0, 1.0e-4)
assert samples.total_steps == 1000
assert samples.total_walkers == 10
assert samples.auto_correlations.times[0] == pytest.approx(
31.98507, 1.0e-4)
def test__non_zero_lower_limit(self):
log_uniform_half = af.LogUniformPrior(lower_limit=0.5, upper_limit=1.0)
assert log_uniform_half.value_for(0.0) == 0.5
assert log_uniform_half.value_for(1.0) == 1.0
assert log_uniform_half.value_for(0.5) == pytest.approx(
0.70710678118, 1.0e-4)
def test__simple_assumptions(self):
log_uniform_simple = af.LogUniformPrior(lower_limit=1.0e-8,
upper_limit=1.0)
assert log_uniform_simple.value_for(0.0) == 1.0e-8
assert log_uniform_simple.value_for(1.0) == 1.0
assert log_uniform_simple.value_for(0.5) == 0.0001
def update_mass_to_light_ratio_prior(
model: af.Model(al.lmp.LightMassProfile),
result: af.Result,
einstein_mass_range: Tuple[float, float],
bins: int = 100,
) -> Optional[af.Model]:
"""
Updates the mass to light ratio parameter of a `LightMassProfile` model (e.g. a bulge or disk) such that the
the `LogUniformPrior` on the mass-to-light ratio of the model-component is set with lower and upper limits that
are a multiple of the Einstein mass computed in the previous SOURCE PIPELINE.
For example, if `einstein_mass_range=[0.01, 5.0]` the mass to light ratio will use priors corresponding to
values which give Einstein masses 1% and 500% of the estimated Einstein mass.
Parameters
----------
model
The light and mass profile whoses priors are passed from the LIGHT PIPELINE.
result
The result of the LIGHT PIPELINE used to pass the priors.
einstein_mass_range
The values a the estimate of the Einstein Mass in the LIGHT PIPELINE is multiplied by to set the lower and upper
limits of the profile's mass-to-light ratio.
bins
The number of bins used to map a calculated Einstein Mass to that of the `LightMassProfile`.
Returns
-------
af.Model(mp.LightMassProfile)
The light and mass profile whose mass-to-light ratio prior is set using the input Einstein mass and range.
"""
if model is None:
return None
grid = result.max_log_likelihood_fit.grid
einstein_radius = result.max_log_likelihood_tracer.einstein_radius_from(grid=grid)
einstein_mass = result.max_log_likelihood_tracer.einstein_mass_angular_from(
grid=grid
)
einstein_mass_lower = einstein_mass_range[0] * einstein_mass
einstein_mass_upper = einstein_mass_range[1] * einstein_mass
instance = model.instance_from_prior_medians()
mass_to_light_ratio_lower = instance.normalization_via_mass_angular_from(
mass_angular=einstein_mass_lower, radius=einstein_radius, bins=bins
)
mass_to_light_ratio_upper = instance.normalization_via_mass_angular_from(
mass_angular=einstein_mass_upper, radius=einstein_radius, bins=bins
)
model.mass_to_light_ratio = af.LogUniformPrior(
lower_limit=mass_to_light_ratio_lower, upper_limit=mass_to_light_ratio_upper
)
return model
def test_change_class():
gaussian_0 = af.Model(af.Gaussian,
normalization=af.UniformPrior(lower_limit=1e-6,
upper_limit=1e6))
gaussian_1 = af.Model(af.Gaussian,
normalization=af.LogUniformPrior(lower_limit=1e-6,
upper_limit=1e6))
assert Identifier(gaussian_0) != Identifier(gaussian_1)
def test_log_prior_list_from_vector(self):
mapper = af.ModelMapper()
mapper.mock_class = af.PriorModel(af.m.MockClassx2)
mapper.mock_class.one = af.GaussianPrior(mean=1.0, sigma=2.0)
mapper.mock_class.two = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=10.0)
log_prior_list = mapper.log_prior_list_from_vector(vector=[0.0, 5.0])
assert log_prior_list == [0.125, 0.2]
def test__samples_from_model(self):
# Setup pickle of mock Dynesty sampler that the samples_from_model function uses.
results = MockDynestyResults(
samples=np.array([[1.0, 2.0, 3.0, 5.0], [1.0, 2.0, 3.0, 4.0],
[1.0, 2.0, 3.0, 4.0]]),
logl=[1.0, 2.0, 3.0],
logwt=[np.log(1.0), np.log(2.0),
np.log(3.0)],
ncall=[5.0, 5.0],
logz=[-2.0, -1.0, 0.0],
nlive=3,
)
sampler = MockDynestySampler(results=results)
paths = af.DirectoryPaths(
path_prefix=path.join("non_linear", "dynesty"))
paths._identifier = "tag"
dynesty = af.DynestyStatic(nlive=3)
dynesty.paths = paths
with open(path.join(dynesty.paths.samples_path, "dynesty.pickle"),
"wb") as f:
dill.dump(sampler, f)
model = af.ModelMapper(mock_class=mock.MockClassx4)
model.mock_class.two = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=10.0)
samples = dynesty.samples_from(model=model)
assert isinstance(samples.parameter_lists, list)
assert isinstance(samples.parameter_lists[0], list)
assert isinstance(samples.log_likelihood_list, list)
assert isinstance(samples.log_prior_list, list)
assert isinstance(samples.log_posterior_list, list)
assert isinstance(samples.weight_list, list)
assert samples.parameter_lists == [
[1.0, 2.0, 3.0, 5.0],
[1.0, 2.0, 3.0, 4.0],
[1.0, 2.0, 3.0, 4.0],
]
assert samples.log_likelihood_list == [1.0, 2.0, 3.0]
assert samples.log_prior_list == [0.2, 0.25, 0.25]
assert samples.weight_list == pytest.approx([1.0, 2.0, 3.0], 1.0e-4)
assert samples.total_samples == 10
assert samples.log_evidence == 0.0
assert samples.number_live_points == 3
def test__median_pdf_parameters(self):
emcee = af.Emcee()
emcee.paths = af.DirectoryPaths(
path_prefix=path.join("non_linear", "emcee"))
emcee.paths._identifier = "tag"
model = af.ModelMapper(mock_class=af.m.MockClassx4)
model.mock_class.two = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=10.0)
samples = emcee.samples_from(model=model)
assert samples.median_pdf_vector == pytest.approx(
[0.008422, -0.026413, 9.9579656, 0.494618], 1.0e-3)
def make_pipeline(name, folders, search=af.DynestyStatic()):
lens = al.GalaxyModel(redshift=0.5, mass=al.mp.EllipticalIsothermal)
lens.mass.centre_0 = af.UniformPrior(lower_limit=-0.01, upper_limit=0.01)
lens.mass.centre_1 = af.UniformPrior(lower_limit=-0.01, upper_limit=0.01)
lens.mass.einstein_radius = af.UniformPrior(lower_limit=1.55,
upper_limit=1.65)
source = al.GalaxyModel(redshift=1.0, light=al.lp.EllipticalSersic)
source.light.centre_0 = af.UniformPrior(lower_limit=-0.01,
upper_limit=0.01)
source.light.centre_1 = af.UniformPrior(lower_limit=-0.01,
upper_limit=0.01)
source.light.intensity = af.UniformPrior(lower_limit=0.35,
upper_limit=0.45)
source.light.effective_radius = af.UniformPrior(lower_limit=0.45,
upper_limit=0.55)
source.light.sersic_index = af.UniformPrior(lower_limit=0.9,
upper_limit=1.1)
class GridPhase(
af.as_grid_search(phase_class=al.PhaseImaging, parallel=True)):
@property
def grid_priors(self):
return [
self.model.galaxies.subhalo.mass.centre_0,
self.model.galaxies.subhalo.mass.centre_1,
]
subhalo = al.GalaxyModel(redshift=0.5,
mass=al.mp.SphericalTruncatedNFWMCRLudlow)
subhalo.mass.mass_at_200 = af.LogUniformPrior(lower_limit=1.0e6,
upper_limit=1.0e11)
subhalo.mass.centre_0 = af.UniformPrior(lower_limit=-2.5, upper_limit=2.5)
subhalo.mass.centre_1 = af.UniformPrior(lower_limit=-2.5, upper_limit=2.5)
phase1 = GridPhase(
phase_name="phase_1",
folders=folders,
galaxies=dict(lens=lens, subhalo=subhalo, source=source),
search=search,
settings=al.SettingsPhaseImaging(),
number_of_steps=2,
)
return al.PipelineDataset(name, phase1)
def test_log_uniform_prior(result):
result.grid_priors = [af.LogUniformPrior()]
assert result.physical_lower_limits_lists == [[1e-06],
[
pytest.approx(0.001,
rel=0.01)
]]
assert result.physical_centres_lists == [[
pytest.approx(3.1622776601683795e-05, rel=0.01)
], [pytest.approx(0.03162277660168379, rel=0.01)]]
assert result.physical_upper_limits_lists == [[
pytest.approx(0.001, rel=0.01)
], [pytest.approx(1.0, rel=0.01)]]
def test__autocorrelation_times(self):
emcee = af.Emcee()
emcee.paths = af.DirectoryPaths(
path_prefix=path.join("non_linear", "emcee"))
emcee.paths._identifier = "tag"
model = af.ModelMapper(mock_class=af.m.MockClassx4)
model.mock_class.two = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=10.0)
samples = emcee.samples_from(model=model)
assert samples.auto_correlations.previous_times == pytest.approx(
[31.1079, 36.0910, 72.44768, 65.86194], 1.0e-4)
assert samples.auto_correlations.times == pytest.approx(
[31.98507, 36.51001, 73.47629, 67.67495], 1.0e-4)
def test_log10(
lower_limit,
upper_limit,
unit
):
prior = af.LogUniformPrior(
lower_limit=lower_limit,
upper_limit=upper_limit
)
assert 10.0 ** (
np.log10(lower_limit)
+ unit * (np.log10(upper_limit) - np.log10(lower_limit))
) == pytest.approx(
prior.value_for(
unit
),
abs=0.001
)
def test__samples_from_model(self, multi_nest_samples_path,
multi_nest_resume_path,
multi_nest_summary_path):
multi_nest = af.MultiNest()
multi_nest.paths = af.DirectoryPaths(
path_prefix=path.join("non_linear", "multinest"))
create_weighted_samples_4_parameters(
file_path=multi_nest.paths.samples_path)
create_resume(file_path=multi_nest.paths.samples_path)
create_summary_4_parameters(file_path=multi_nest.paths.samples_path)
model = af.ModelMapper(mock_class=mock.MockClassx4)
model.mock_class.two = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=10.0)
samples = multi_nest.samples_from(model=model)
assert samples.parameter_lists == [
[1.1, 2.1, 3.1, 4.1],
[0.9, 1.9, 2.9, 3.9],
[1.0, 2.0, 3.0, 4.0],
[1.0, 2.0, 3.0, 4.0],
[1.0, 2.0, 3.0, 4.0],
[1.0, 2.0, 3.0, 4.0],
[1.0, 2.0, 3.0, 4.0],
[1.0, 2.0, 3.0, 4.0],
[1.0, 2.0, 3.0, 4.0],
[1.0, 2.0, 3.0, 4.0],
]
value = -0.5 * 9999999.9
assert samples.log_likelihood_list == 10 * [value]
assert samples.log_prior_list == pytest.approx([
0.243902, 0.256410, 0.25, 0.25, 0.25, 0.25, 0.25, 0.25, 0.25, 0.25
], 1.0e-4)
assert samples.weight_list == [
0.02, 0.02, 0.01, 0.05, 0.1, 0.1, 0.1, 0.1, 0.2, 0.3
]
assert samples.total_samples == 12345
assert samples.log_evidence == 0.02
assert samples.number_live_points == 50
def customize_priors(self, results):
### Lens Mass, PL -> PL, Shear -> Shear ###
# self.galaxies.lens = results.from_phase('phase_1').\
# constant.galaxies.lens
### Lens Subhalo, Adjust priors to physical masses (10^6 - 10^10) and concentrations (6-24)
self.galaxies.subhalo.mass.mass_at_200 = af.LogUniformPrior(
lower_limit=10.0e6, upper_limit=10.0e9)
self.galaxies.subhalo.mass.centre_0 = af.UniformPrior(
lower_limit=-2.0, upper_limit=2.0)
self.galaxies.subhalo.mass.centre_1 = af.UniformPrior(
lower_limit=-2.0, upper_limit=2.0)
### Source Light, Sersic -> Sersic ###
self.galaxies.source.light.centre = (
results.from_phase("phase_1").variable_absolute(
a=0.05).galaxies.source.light.centre)
self.galaxies.source.light.intensity = (
results.from_phase("phase_1").variable_relative(
r=0.5).galaxies.source.light.intensity)
self.galaxies.source.light.effective_radius = (
results.from_phase("phase_1").variable_relative(
r=0.5).galaxies.source.light.effective_radius)
self.galaxies.source.light.sersic_index = (
results.from_phase("phase_1").variable_relative(
r=0.5).galaxies.source.light.sersic_index)
self.galaxies.source.light.axis_ratio = (
results.from_phase("phase_1").variable_absolute(
a=0.1).galaxies.source.light.axis_ratio)
self.galaxies.source.light.phi = (
results.from_phase("phase_1").variable_absolute(
a=20.0).galaxies.source.light.phi)
def test__vector_at_sigma__uses_output_files(self):
emcee = af.Emcee()
emcee.paths = af.DirectoryPaths(
path_prefix=path.join("non_linear", "emcee"))
emcee.paths._identifier = "tag"
model = af.ModelMapper(mock_class=af.m.MockClassx4)
model.mock_class.two = af.LogUniformPrior(lower_limit=1e-8,
upper_limit=10.0)
samples = emcee.samples_from(model=model)
parameters = samples.vector_at_sigma(sigma=3.0)
assert parameters[0][0:2] == pytest.approx((-0.003197, 0.019923), 1e-2)
parameters = samples.vector_at_sigma(sigma=1.0)
assert parameters[0][0:2] == pytest.approx((0.0042278, 0.01087681),
1e-2)
def hyper_galaxy_via_galaxy_model_from(self,
galaxy_model,
galaxy_instance,
noise_factor_is_model=False):
hyper_galaxy = af.Model(ag.HyperGalaxy)
if galaxy_model.hyper_galaxy is None:
return None
if not noise_factor_is_model:
hyper_galaxy.noise_factor = galaxy_instance.hyper_galaxy.noise_factor
else:
hyper_galaxy.noise_factor = af.LogUniformPrior(
lower_limit=1e-4,
upper_limit=2.0 * galaxy_instance.hyper_galaxy.noise_factor,
)
hyper_galaxy.contribution_factor = (
galaxy_instance.hyper_galaxy.contribution_factor)
hyper_galaxy.noise_power = galaxy_instance.hyper_galaxy.noise_power
return hyper_galaxy
color="k",
ecolor="k",
elinewidth=1,
capsize=2,
)
plt.show()
plt.close()
"""
__Model + Analysis__
We create the model and analysis, which in this example is a single `Gaussian` and therefore has dimensionality N=3.
"""
model = af.Model(m.Gaussian)
model.centre = af.UniformPrior(lower_limit=0.0, upper_limit=100.0)
model.intensity = af.LogUniformPrior(lower_limit=1e-2, upper_limit=1e2)
model.sigma = af.UniformPrior(lower_limit=0.0, upper_limit=30.0)
analysis = a.Analysis(data=data, noise_map=noise_map)
"""
__Search__
We now create and run the `MultiNest` object which acts as our non-linear search.
We manually specify all of the MultiNest settings, descriptions of which are provided at the following webpage:
- https://github.com/JohannesBuchner/MultiNest
- https://github.com/JohannesBuchner/PyMultiNest
- http://johannesbuchner.github.io/PyMultiNest/index.html#
"""
multi_nest = af.MultiNest(
visibilities=np.add(visibilities, noise_map),
noise_map=noise_map,
z_step_kms=z_step_kms)
lens_model = af.PriorModel(mass_profiles.EllipticalPowerLaw)
lens_model.centre_0 = af.GaussianPrior(mean=0.0, sigma=0.25)
lens_model.centre_1 = af.GaussianPrior(mean=0.0, sigma=0.25)
lens_model.einstein_radius = af.UniformPrior(lower_limit=0.85,
upper_limit=1.25)
lens_model.slope = 2.0
src_model = af.PriorModel(profiles.Kinematical)
src_model.centre_0 = af.GaussianPrior(mean=0.0, sigma=0.25)
src_model.centre_1 = af.GaussianPrior(mean=0.0, sigma=0.25)
src_model.z_centre = af.GaussianPrior(mean=16.0, sigma=2.0)
src_model.intensity = af.LogUniformPrior(lower_limit=10**-2.0,
upper_limit=10**+2.0)
src_model.maximum_velocity = af.UniformPrior(lower_limit=25.0,
upper_limit=400.0)
src_model.velocity_dispersion = af.UniformPrior(lower_limit=0.0,
upper_limit=100.0)
phase_folders = [
string_utils.remove_substring_from_end_of_string(
string=os.path.basename(__file__), substring=".py")
]
phase_1 = phase.Phase(
phase_name="phase_1__version_{}".format(autolens_version),
phase_folders=phase_folders,
profiles=af.CollectionPriorModel(
lens=lens_model,
# ncols=8,
# cube_contours=lensed_cube,
# )
# exit()
lens_model = af.PriorModel(mass_profiles.EllipticalPowerLaw)
lens_model.centre_0 = af.GaussianPrior(mean=0.0, sigma=0.25)
lens_model.centre_1 = af.GaussianPrior(mean=0.0, sigma=0.25)
lens_model.einstein_radius = af.UniformPrior(lower_limit=0.85,
upper_limit=1.25)
lens_model.slope = 2.0
source_model_1 = af.PriorModel(profiles.EllipticalSersic)
source_model_1.centre_0 = af.GaussianPrior(mean=0.0, sigma=0.25)
source_model_1.centre_1 = af.GaussianPrior(mean=0.0, sigma=0.25)
source_model_1.intensity = af.LogUniformPrior(lower_limit=5.0 * 10**-6.0,
upper_limit=5.0 * 10**-4.0)
source_model_2 = af.PriorModel(profiles.Kinematical)
source_model_2.centre_0 = af.GaussianPrior(mean=0.0, sigma=0.25)
source_model_2.centre_1 = af.GaussianPrior(mean=0.0, sigma=0.25)
source_model_2.z_centre = af.GaussianPrior(mean=16.0, sigma=2.0)
source_model_2.intensity = af.LogUniformPrior(lower_limit=10**-2.0,
upper_limit=10**+2.0)
source_model_2.maximum_velocity = af.UniformPrior(lower_limit=25.0,
upper_limit=400.0)
source_model_2.velocity_dispersion = af.UniformPrior(lower_limit=0.0,
upper_limit=100.0)
phase_folders = [
string_utils.remove_substring_from_end_of_string(
string=os.path.basename(__file__), substring=".py")
def test__identifier_description__after_take_attributes():
model = af.CollectionPriorModel(gaussian=af.PriorModel(
af.Gaussian,
centre=af.UniformPrior(lower_limit=0.0, upper_limit=1.0),
normalization=af.LogUniformPrior(lower_limit=0.001, upper_limit=0.01),
sigma=af.GaussianPrior(
mean=0.5, sigma=2.0, lower_limit=-1.0, upper_limit=1.0),
))
model.take_attributes(source=model)
identifier = Identifier([model])
description = identifier.description.splitlines()
# THIS TEST FAILS DUE TO THE BUG DESCRIBED IN A GITHUB ISSUE.
i = 0
assert description[i] == "CollectionPriorModel"
i += 1
assert description[i] == "item_number"
i += 1
assert description[i] == "0"
i += 1
assert description[i] == "gaussian"
i += 1
assert description[i] == "PriorModel"
i += 1
assert description[i] == "cls"
i += 1
assert description[i] == "autofit.example.model.Gaussian"
i += 1
assert description[i] == "centre"
i += 1
assert description[i] == "UniformPrior"
i += 1
assert description[i] == "lower_limit"
i += 1
assert description[i] == "0.0"
i += 1
assert description[i] == "upper_limit"
i += 1
assert description[i] == "1.0"
i += 1
assert description[i] == "normalization"
i += 1
assert description[i] == "LogUniformPrior"
i += 1
assert description[i] == "lower_limit"
i += 1
assert description[i] == "0.001"
i += 1
assert description[i] == "upper_limit"
i += 1
assert description[i] == "0.01"
i += 1
assert description[i] == "sigma"
i += 1
assert description[i] == "GaussianPrior"
i += 1
assert description[i] == "lower_limit"
i += 1
assert description[i] == "-1.0"
i += 1
assert description[i] == "upper_limit"
i += 1
assert description[i] == "1.0"
i += 1
assert description[i] == "mean"
i += 1
assert description[i] == "0.5"
i += 1
assert description[i] == "sigma"
i += 1
assert description[i] == "2.0"
i += 1
def test__identifier_description__after_model_and_instance():
model = af.CollectionPriorModel(gaussian=af.PriorModel(
af.Gaussian,
centre=af.UniformPrior(lower_limit=0.0, upper_limit=1.0),
normalization=af.LogUniformPrior(lower_limit=0.001, upper_limit=0.01),
sigma=af.GaussianPrior(
mean=0.5, sigma=2.0, lower_limit=-1.0, upper_limit=1.0),
))
max_log_likelihood_instance = model.instance_from_prior_medians()
samples = af.m.MockSamples(
max_log_likelihood_instance=max_log_likelihood_instance,
gaussian_tuples=[(1.0, 2.0), (3.0, 4.0), (5.0, 6.0)])
search = af.m.MockSearch(prior_passer=af.PriorPasser(
sigma=1.0, use_errors=True, use_widths=False))
result = af.Result(samples=samples, model=model, search=search)
model.gaussian.centre = result.model.gaussian.centre
model.gaussian.normalization = result.instance.gaussian.normalization
identifier = Identifier([model])
description = identifier.description.splitlines()
# THIS TEST FAILS DUE TO THE BUG DESCRIBED IN A GITHUB ISSUE.
i = 0
assert description[i] == "CollectionPriorModel"
i += 1
assert description[i] == "item_number"
i += 1
assert description[i] == "0"
i += 1
assert description[i] == "gaussian"
i += 1
assert description[i] == "PriorModel"
i += 1
assert description[i] == "cls"
i += 1
assert description[i] == "autofit.example.model.Gaussian"
i += 1
assert description[i] == "centre"
i += 1
assert description[i] == "GaussianPrior"
i += 1
assert description[i] == "lower_limit"
i += 1
assert description[i] == "0.0"
i += 1
assert description[i] == "upper_limit"
i += 1
assert description[i] == "1.0"
i += 1
assert description[i] == "mean"
i += 1
assert description[i] == "1.0"
i += 1
assert description[i] == "sigma"
i += 1
assert description[i] == "2.0"
i += 1
assert description[i] == "normalization"
i += 1
assert description[i] == "0.00316228"
i += 1
assert description[i] == "sigma"
i += 1
assert description[i] == "GaussianPrior"
i += 1
assert description[i] == "lower_limit"
i += 1
assert description[i] == "-1.0"
i += 1
assert description[i] == "upper_limit"
i += 1
assert description[i] == "1.0"
i += 1
assert description[i] == "mean"
i += 1
assert description[i] == "0.5"
i += 1
assert description[i] == "sigma"
i += 1
assert description[i] == "2.0"
i += 1
def test__identifier_description():
model = af.CollectionPriorModel(gaussian=af.PriorModel(
af.Gaussian,
centre=af.UniformPrior(lower_limit=0.0, upper_limit=1.0),
normalization=af.LogUniformPrior(lower_limit=0.001, upper_limit=0.01),
sigma=af.GaussianPrior(
mean=0.5, sigma=2.0, lower_limit=-1.0, upper_limit=1.0),
))
identifier = Identifier([model])
description = identifier.description.splitlines()
i = 0
assert description[i] == "CollectionPriorModel"
i += 1
assert description[i] == "item_number"
i += 1
assert description[i] == "0"
i += 1
assert description[i] == "gaussian"
i += 1
assert description[i] == "PriorModel"
i += 1
assert description[i] == "cls"
i += 1
assert description[i] == "autofit.example.model.Gaussian"
i += 1
assert description[i] == "centre"
i += 1
assert description[i] == "UniformPrior"
i += 1
assert description[i] == "lower_limit"
i += 1
assert description[i] == "0.0"
i += 1
assert description[i] == "upper_limit"
i += 1
assert description[i] == "1.0"
i += 1
assert description[i] == "normalization"
i += 1
assert description[i] == "LogUniformPrior"
i += 1
assert description[i] == "lower_limit"
i += 1
assert description[i] == "0.001"
i += 1
assert description[i] == "upper_limit"
i += 1
assert description[i] == "0.01"
i += 1
assert description[i] == "sigma"
i += 1
assert description[i] == "GaussianPrior"
i += 1
assert description[i] == "lower_limit"
i += 1
assert description[i] == "-1.0"
i += 1
assert description[i] == "upper_limit"
i += 1
assert description[i] == "1.0"
i += 1
assert description[i] == "mean"
i += 1
assert description[i] == "0.5"
i += 1
assert description[i] == "sigma"
i += 1
assert description[i] == "2.0"
i += 1
def make_pipeline(
setup,
phase_folders,
real_space_mask,
lens_redshift,
source_redshift,
instance,
priors=None,
pipeline_name="pipeline_source__inversion",
transformer_class=al.TransformerNUFFT,
auto_positions_factor=None,
positions_threshold=None,
sub_size=1,
inversion_uses_border=True,
inversion_pixel_limit=None,
evidence_tolerance=10.0,
):
phase_folders.append(pipeline_name)
for type in ["general", "source"]:
if hasattr(setup, type):
if type == "general":
phase_folders.append(setup.general.source_tag)
if type == "source":
setup.set_source_type(source_type=setup.source.inversion_tag)
phase_folders.append(setup.source.tag)
if "lens" in instance.keys():
lens = instance["lens"]
else:
raise ValueError("...")
source = al.GalaxyModel(
redshift=source_redshift,
pixelization=al.pix.VoronoiMagnification,
regularization=al.reg.Constant,
)
source.pixelization.shape.shape_0 = af.UniformPrior(lower_limit=5,
upper_limit=100)
source.pixelization.shape.shape_1 = af.UniformPrior(lower_limit=5,
upper_limit=100)
source.regularization.coefficient = af.LogUniformPrior(
lower_limit=10**-5.0, upper_limit=10**+5.0)
phase1 = al.PhaseInterferometer(
phase_name="phase_1__lens_instance__source_inversion",
phase_folders=phase_folders,
real_space_mask=real_space_mask,
galaxies=dict(
lens=lens,
source=source,
),
transformer_class=transformer_class,
positions_threshold=positions_threshold,
auto_positions_factor=auto_positions_factor,
sub_size=sub_size,
inversion_uses_border=inversion_uses_border,
inversion_pixel_limit=inversion_pixel_limit,
non_linear_class=af.MultiNest,
)
phase1.optimizer.const_efficiency_mode = True
phase1.optimizer.n_live_points = 20
phase1.optimizer.sampling_efficiency = 0.8
phase1.optimizer.evidence_tolerance = 0.1
return al.PipelineDataset(pipeline_name, phase1)
source_1 = al.GalaxyModel(
redshift=source_redshift,
pixelization=al.pix.VoronoiMagnification,
regularization=al.reg.Constant,
)
source_2 = al.GalaxyModel(
redshift=source_redshift,
pixelization=al.pix.VoronoiMagnification,
regularization=al.reg.Constant,
)
source_1.pixelization.shape.shape_0 = af.UniformPrior(lower_limit=5,
upper_limit=50)
source_1.pixelization.shape.shape_1 = af.UniformPrior(lower_limit=5,
upper_limit=50)
source_1.regularization.coefficient = af.LogUniformPrior(
lower_limit=10**-6.0, upper_limit=10**8.0)
source_2.pixelization.shape = source_1.pixelization.shape
source_2.regularization.coefficient = af.LogUniformPrior(
lower_limit=10**-6.0, upper_limit=10**8.0)
# source_1.pixelization.shape = (15, 15)
# source_1.regularization.coefficient = 5000.0
# source_2.pixelization.shape = (15, 15)
# source_2.regularization.coefficient = 1000000.0
lens.mass.centre_0 = 0.0
lens.mass.centre_1 = 0.0
lens.mass.axis_ratio = 0.75
lens.mass.phi = 45.0
lens.mass.einstein_radius = 1.0
def make_pipeline(phase_folders=None):
### SETUP PIPELINE AND PHASE NAMES, TAGS AND PATHS ###
# We setup the pipeline name using the tagging module. In this case, the pipeline name is not given a tag and
# will be the string specified below However, its good practise to use the 'tag.' function below, incase
# a pipeline does use customized tag names.
pipeline_name = "pipeline__inversion"
# This function uses the phase folders and pipeline name to set up the output directory structure,
# e.g. 'autolens_workspace/output/phase_folder_1/phase_folder_2/pipeline_name/phase_name/'
phase_folders.append(pipeline_name)
# This is the same phase 1 as the complex source pipeline, which we saw gave a good fit to the overall
# structure of the lensed source and provided an accurate lens mass model.
phase1 = al.PhaseImaging(
phase_name="phase_1__lens_sie__source_sersic",
phase_folders=phase_folders,
galaxies=dict(
lens=al.GalaxyModel(redshift=0.5, mass=al.mp.EllipticalIsothermal),
source=al.GalaxyModel(redshift=1.0, light=al.lp.EllipticalSersic),
),
optimizer_class=af.MultiNest,
)
phase1.optimizer.sampling_efficiency = 0.3
phase1.optimizer.const_efficiency_mode = True
# Now, in phase 2, lets use the lens mass model to fit the source with an inversion.
source = al.GalaxyModel(
redshift=1.0,
pixelization=al.pix.VoronoiMagnification,
regularization=al.reg.Constant,
)
# We can customize the inversion's priors like we do our light and mass profiles.
source.pixelization.shape_0 = af.UniformPrior(lower_limit=20.0,
upper_limit=40.0)
source.pixelization.shape_1 = af.UniformPrior(lower_limit=20.0,
upper_limit=40.0)
# The expected value of the regularization coefficient depends on the details of the dataset reduction and
# source galaxy. A broad log-uniform prior is thus an appropriate way to sample the large range of
# possible values.
source.regularization.coefficient = af.LogUniformPrior(lower_limit=1.0e-6,
upper_limit=10000.0)
phase2 = al.PhaseImaging(
phase_name="phase_2__source_inversion_initialize",
phase_folders=phase_folders,
galaxies=dict(
lens=al.GalaxyModel(redshift=0.5,
mass=phase1.result.model.galaxies.lens.mass),
source=source,
),
optimizer_class=af.MultiNest,
)
phase2.optimizer.sampling_efficiency = 0.3
phase2.optimizer.const_efficiency_mode = True
# We now 'extend' phase 1 with an additional 'inversion phase' which uses the best-fit mass model of phase 1 above
# to refine the it inversion used, by fitting only the pixelization & regularization parameters.
# The the inversion phase results are accessible as attributes of the phase results and used in phase 3 below.
phase2 = phase2.extend_with_inversion_phase()
# Now, in phase 3, lets use the refined source inversion to fit the lens mass model again.
phase3 = al.PhaseImaging(
phase_name="phase_3__lens_sie__source_inversion",
phase_folders=phase_folders,
galaxies=dict(
lens=al.GalaxyModel(redshift=0.5,
mass=phase1.result.model.galaxies.lens.mass),
source=al.GalaxyModel(
redshift=1.0,
pixelization=phase2.inversion.instance.galaxies.source.
pixelization,
regularization=phase2.inversion.instance.galaxies.source.
regularization,
),
),
optimizer_class=af.MultiNest,
)
phase3.optimizer.sampling_efficiency = 0.3
phase3.optimizer.const_efficiency_mode = True
return al.PipelineDataset(pipeline_name, phase1, phase2, phase3)
