def positions_fit_for_tracer(self, tracer, hyper_image_sky,
hyper_background_noise):
return fit.FitImaging(
masked_imaging=self.masked_dataset,
tracer=tracer,
hyper_image_sky=hyper_image_sky,
hyper_background_noise=hyper_background_noise,
)
def masked_imaging_fit_for_tracer(
self, tracer, hyper_image_sky, hyper_background_noise
):
return fit.FitImaging(
masked_imaging=self.masked_dataset,
tracer=tracer,
hyper_image_sky=hyper_image_sky,
hyper_background_noise=hyper_background_noise,
settings_pixelization=self.settings.settings_pixelization,
settings_inversion=self.settings.settings_inversion,
)
def stochastic_log_evidences_for_instance(self, instance):
instance = self.associate_hyper_images(instance=instance)
tracer = self.tracer_for_instance(instance=instance)
if not tracer.has_pixelization:
return
if not isinstance(
tracer.pixelizations_of_planes[-1], pix.VoronoiBrightnessImage
):
return
hyper_image_sky = self.hyper_image_sky_for_instance(instance=instance)
hyper_background_noise = self.hyper_background_noise_for_instance(
instance=instance
)
settings_pixelization = (
self.settings.settings_pixelization.settings_with_is_stochastic_true()
)
log_evidences = []
for i in range(self.settings.settings_lens.stochastic_samples):
try:
log_evidence = fit.FitImaging(
masked_imaging=self.masked_dataset,
tracer=tracer,
hyper_image_sky=hyper_image_sky,
hyper_background_noise=hyper_background_noise,
settings_pixelization=settings_pixelization,
settings_inversion=self.settings.settings_inversion,
).log_evidence
except (
PixelizationException,
InversionException,
GridException,
OverflowError,
) as e:
log_evidence = None
if log_evidence is not None:
log_evidences.append(log_evidence)
return log_evidences
def masked_imaging_fit_for_tracer(self,
tracer,
hyper_image_sky,
hyper_background_noise,
use_hyper_scalings=True):
return fit.FitImaging(
masked_imaging=self.masked_dataset,
tracer=tracer,
hyper_image_sky=hyper_image_sky,
hyper_background_noise=hyper_background_noise,
use_hyper_scaling=use_hyper_scalings,
settings_pixelization=self.settings.settings_pixelization,
settings_inversion=self.settings.settings_inversion,
preloads=self.preloads,
)
def log_likelihood_function(self, instance):
"""
Determine the fit of a lens galaxy and source galaxy to the masked_imaging in this lens.
Parameters
----------
instance
A model instance with attributes
Returns
-------
fit : Fit
A fractional value indicating how well this model fit and the model masked_imaging itself
"""
self.associate_hyper_images(instance=instance)
tracer = self.tracer_for_instance(instance=instance)
self.settings.settings_lens.check_positions_trace_within_threshold_via_tracer(
tracer=tracer, positions=self.masked_dataset.positions)
hyper_image_sky = self.hyper_image_sky_for_instance(instance=instance)
hyper_background_noise = self.hyper_background_noise_for_instance(
instance=instance)
if self.settings.settings_lens.stochastic_likelihood_resamples is None:
try:
return self.masked_imaging_fit_for_tracer(
tracer=tracer,
hyper_image_sky=hyper_image_sky,
hyper_background_noise=hyper_background_noise,
).figure_of_merit
except (
PixelizationException,
InversionException,
GridException,
OverflowError,
) as e:
raise FitException from e
else:
figures_of_merit = []
for i in range(self.settings.settings_lens.
stochastic_likelihood_resamples):
settings_pixelization = copy.deepcopy(
self.settings.settings_pixelization)
settings_pixelization.kmeans_seed = i
# settings_pixelization.is_stochastic = True
try:
figures_of_merit.append(
fit.FitImaging(
masked_imaging=self.masked_dataset,
tracer=tracer,
hyper_image_sky=hyper_image_sky,
hyper_background_noise=hyper_background_noise,
settings_pixelization=settings_pixelization,
settings_inversion=self.settings.
settings_inversion,
).log_evidence)
except (
PixelizationException,
InversionException,
GridException,
OverflowError,
) as e:
raise FitException from e
return np.mean(figures_of_merit)