Skip to content
Snippets Groups Projects
Commit 9aa84e19 authored by Gregory Ashton's avatar Gregory Ashton
Browse files

Merge branch 'refactor-likelihood' into 'master' 

minor refactoring of GWT likelihood

See merge request !495
parents 9f8ec3b1 c830634a
No related branches found
No related tags found
1 merge request!495minor refactoring of GWT likelihood
Pipeline #63802 passed
Stage: test
Stage: deploy
Hide whitespace changes
Inline Side-by-side
Showing
with 71 additions and 65 deletions
bilby/gw/likelihood.py
+ 71
− 65
View file @ 9aa84e19
......@@ -189,17 +189,17 @@ class GravitationalWaveTransient(likelihood.Likelihood):
@property
def priors(self):
return self.__prior
return self._prior
@priors.setter
def priors(self, priors):
if priors is not None:
self.__prior = priors.copy()
self._prior = priors.copy()
elif any([self.time_marginalization, self.phase_marginalization,
self.distance_marginalization]):
raise ValueError("You can't use a marginalized likelihood without specifying a priors")
else:
self.__prior = None
self._prior = None
def noise_log_likelihood(self):
log_l = 0
......@@ -210,7 +210,7 @@ class GravitationalWaveTransient(likelihood.Likelihood):
interferometer.frequency_domain_strain[mask],
interferometer.power_spectral_density_array[mask],
self.waveform_generator.duration) / 2
return log_l.real
return float(np.real(log_l))
def log_likelihood_ratio(self):
waveform_polarizations =\
......@@ -222,57 +222,40 @@ class GravitationalWaveTransient(likelihood.Likelihood):
d_inner_h = 0.
optimal_snr_squared = 0.
complex_matched_filter_snr = 0.
d_inner_h_squared_tc_array = np.zeros(
self.interferometers.frequency_array[0:-1].shape,
dtype=np.complex128)
if self.time_marginalization:
d_inner_h_tc_array = np.zeros(
self.interferometers.frequency_array[0:-1].shape,
dtype=np.complex128)
for interferometer in self.interferometers:
per_detector_snr = self.calculate_snrs(
waveform_polarizations, interferometer)
waveform_polarizations=waveform_polarizations,
interferometer=interferometer)
d_inner_h += per_detector_snr.d_inner_h
optimal_snr_squared += per_detector_snr.optimal_snr_squared
optimal_snr_squared += np.real(per_detector_snr.optimal_snr_squared)
complex_matched_filter_snr += per_detector_snr.complex_matched_filter_snr
if self.time_marginalization:
d_inner_h_squared_tc_array += per_detector_snr.d_inner_h_squared_tc_array
d_inner_h_tc_array += per_detector_snr.d_inner_h_squared_tc_array
if self.time_marginalization:
if self.distance_marginalization:
rho_mf_ref_tc_array, rho_opt_ref = self._setup_rho(
d_inner_h_squared_tc_array, optimal_snr_squared)
if self.phase_marginalization:
dist_marged_log_l_tc_array = self._interp_dist_margd_loglikelihood(
abs(rho_mf_ref_tc_array), rho_opt_ref)
else:
dist_marged_log_l_tc_array = self._interp_dist_margd_loglikelihood(
rho_mf_ref_tc_array.real, rho_opt_ref)
log_l = logsumexp(dist_marged_log_l_tc_array,
b=self.time_prior_array)
elif self.phase_marginalization:
log_l = logsumexp(self._bessel_function_interped(abs(
d_inner_h_squared_tc_array)),
b=self.time_prior_array) - optimal_snr_squared / 2
else:
log_l = logsumexp(
d_inner_h_squared_tc_array.real,
b=self.time_prior_array) - optimal_snr_squared / 2
log_l = self.time_marginalized_likelihood(
d_inner_h_tc_array=d_inner_h_tc_array,
h_inner_h=optimal_snr_squared)
elif self.distance_marginalization:
rho_mf_ref, rho_opt_ref = self._setup_rho(d_inner_h, optimal_snr_squared)
if self.phase_marginalization:
rho_mf_ref = abs(rho_mf_ref)
log_l = self._interp_dist_margd_loglikelihood(rho_mf_ref.real, rho_opt_ref)[0]
log_l = self.distance_marginalized_likelihood(
d_inner_h=d_inner_h, h_inner_h=optimal_snr_squared)
elif self.phase_marginalization:
d_inner_h = self._bessel_function_interped(abs(d_inner_h))
log_l = d_inner_h - optimal_snr_squared / 2
log_l = self.phase_marginalized_likelihood(
d_inner_h=d_inner_h, h_inner_h=optimal_snr_squared)
else:
log_l = d_inner_h.real - optimal_snr_squared / 2
log_l = np.real(d_inner_h) - optimal_snr_squared / 2
return log_l.real
return float(log_l.real)
def generate_posterior_sample_from_marginalized_likelihood(self):
"""
......@@ -352,14 +335,8 @@ class GravitationalWaveTransient(likelihood.Likelihood):
h_inner_h += ifo.optimal_snr_squared(signal=signal).real
if self.distance_marginalization:
rho_mf_ref_tc_array, rho_opt_ref = self._setup_rho(
time_log_like = self.distance_marginalized_likelihood(
d_inner_h, h_inner_h)
if self.phase_marginalization:
time_log_like = self._interp_dist_margd_loglikelihood(
abs(rho_mf_ref_tc_array), rho_opt_ref)
else:
time_log_like = self._interp_dist_margd_loglikelihood(
rho_mf_ref_tc_array.real, rho_opt_ref)
elif self.phase_marginalization:
time_log_like = (self._bessel_function_interped(abs(d_inner_h)) -
h_inner_h.real / 2)
......@@ -479,13 +456,39 @@ class GravitationalWaveTransient(likelihood.Likelihood):
new_phase = Interped(phases, phase_post).sample()
return new_phase
def distance_marginalized_likelihood(self, d_inner_h, h_inner_h):
d_inner_h_ref, h_inner_h_ref = self._setup_rho(
d_inner_h, h_inner_h)
if self.phase_marginalization:
d_inner_h_ref = np.abs(d_inner_h_ref)
else:
d_inner_h_ref = np.real(d_inner_h_ref)
return self._interp_dist_margd_loglikelihood(
d_inner_h_ref, h_inner_h_ref)
def phase_marginalized_likelihood(self, d_inner_h, h_inner_h):
d_inner_h = self._bessel_function_interped(abs(d_inner_h))
return d_inner_h - h_inner_h / 2
def time_marginalized_likelihood(self, d_inner_h_tc_array, h_inner_h):
if self.distance_marginalization:
log_l_tc_array = self.distance_marginalized_likelihood(
d_inner_h=d_inner_h_tc_array, h_inner_h=h_inner_h)
elif self.phase_marginalization:
log_l_tc_array = self.phase_marginalized_likelihood(
d_inner_h=d_inner_h_tc_array,
h_inner_h=h_inner_h)
else:
log_l_tc_array = np.real(d_inner_h_tc_array) - h_inner_h / 2
return logsumexp(log_l_tc_array, b=self.time_prior_array)
def _setup_rho(self, d_inner_h, optimal_snr_squared):
rho_opt_ref = (optimal_snr_squared.real *
self.parameters['luminosity_distance'] ** 2 /
self._ref_dist ** 2.)
rho_mf_ref = (d_inner_h * self.parameters['luminosity_distance'] /
self._ref_dist)
return rho_mf_ref, rho_opt_ref
optimal_snr_squared_ref = (optimal_snr_squared.real *
self.parameters['luminosity_distance'] ** 2 /
self._ref_dist ** 2.)
d_inner_h_ref = (d_inner_h * self.parameters['luminosity_distance'] /
self._ref_dist)
return d_inner_h_ref, optimal_snr_squared_ref
def log_likelihood(self):
return self.log_likelihood_ratio() + self.noise_log_likelihood()
......@@ -500,12 +503,12 @@ class GravitationalWaveTransient(likelihood.Likelihood):
return self._distance_array[0]
@property
def _rho_opt_ref_array(self):
def _optimal_snr_squared_ref_array(self):
""" Optimal filter snr at fiducial distance of ref_dist Mpc """
return np.logspace(-5, 10, self._dist_margd_loglikelihood_array.shape[0])
@property
def _rho_mf_ref_array(self):
def _d_inner_h_ref_array(self):
""" Matched filter snr at fiducial distance of ref_dist Mpc """
if self.phase_marginalization:
return np.logspace(-5, 10, self._dist_margd_loglikelihood_array.shape[1])
......@@ -527,7 +530,7 @@ class GravitationalWaveTransient(likelihood.Likelihood):
else:
self._create_lookup_table()
self._interp_dist_margd_loglikelihood = UnsortedInterp2d(
self._rho_mf_ref_array, self._rho_opt_ref_array,
self._d_inner_h_ref_array, self._optimal_snr_squared_ref_array,
self._dist_margd_loglikelihood_array)
@property
......@@ -594,17 +597,20 @@ class GravitationalWaveTransient(likelihood.Likelihood):
logger.info('Building lookup table for distance marginalisation.')
self._dist_margd_loglikelihood_array = np.zeros((400, 800))
for ii, rho_opt_ref in enumerate(self._rho_opt_ref_array):
for jj, rho_mf_ref in enumerate(self._rho_mf_ref_array):
optimal_snr_squared_array = rho_opt_ref * self._ref_dist ** 2. / self._distance_array ** 2
d_inner_h_array = rho_mf_ref * self._ref_dist / self._distance_array
for ii, optimal_snr_squared_ref in enumerate(self._optimal_snr_squared_ref_array):
for jj, d_inner_h_ref in enumerate(self._d_inner_h_ref_array):
optimal_snr_squared_array = (
optimal_snr_squared_ref * self._ref_dist ** 2. /
self._distance_array ** 2)
d_inner_h_array = (
d_inner_h_ref * self._ref_dist / self._distance_array)
if self.phase_marginalization:
d_inner_h_array =\
self._bessel_function_interped(abs(d_inner_h_array))
self._dist_margd_loglikelihood_array[ii][jj] = \
logsumexp(d_inner_h_array - optimal_snr_squared_array / 2,
b=self.distance_prior_array * self._delta_distance)
log_norm = logsumexp(0. / self._distance_array - 0. / self._distance_array ** 2.,
log_norm = logsumexp(0. / self._distance_array,
b=self.distance_prior_array * self._delta_distance)
self._dist_margd_loglikelihood_array -= log_norm
self.cache_lookup_table()
......@@ -816,13 +822,13 @@ class ROQGravitationalWaveTransient(GravitationalWaveTransient):
self.frequency_nodes_quadratic = \
waveform_generator.waveform_arguments['frequency_nodes_quadratic']
def calculate_snrs(self, signal, interferometer):
def calculate_snrs(self, waveform_polarizations, interferometer):
"""
Compute the snrs for ROQ
Parameters
----------
signal: waveform
waveform_polarizations: waveform
interferometer: bilby.gw.detector.Interferometer
"""
......@@ -847,12 +853,12 @@ class ROQGravitationalWaveTransient(GravitationalWaveTransient):
self.frequency_nodes_quadratic,
prefix='recalib_{}_'.format(interferometer.name), **self.parameters)
h_plus_linear = f_plus * signal['linear']['plus'] * calib_linear
h_cross_linear = f_cross * signal['linear']['cross'] * calib_linear
h_plus_linear = f_plus * waveform_polarizations['linear']['plus'] * calib_linear
h_cross_linear = f_cross * waveform_polarizations['linear']['cross'] * calib_linear
h_plus_quadratic = (
f_plus * signal['quadratic']['plus'] * calib_quadratic)
f_plus * waveform_polarizations['quadratic']['plus'] * calib_quadratic)
h_cross_quadratic = (
f_cross * signal['quadratic']['cross'] * calib_quadratic)
f_cross * waveform_polarizations['quadratic']['cross'] * calib_quadratic)
indices, in_bounds = self._closest_time_indices(
ifo_time, self.weights['time_samples'])
......
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment