Skip to content

Commit

Permalink
Add calculation of source type probabilities to PyCBC Live (gwastro#3077
Browse files Browse the repository at this point in the history 
)

* Introduce source probability calculation

* Mass gap probabilities improved

* Change save function to add probs file

* Change in z_delta and mchirp_delta

* Add argument for mchirp_delta coefficient

* Add mchirp bbh condition

* Change save function

* Add CBC probabilities to upload method

* Small corrections to mchirp_area.py

* Add logging messages

* Fix small errors

* Fix small syntax error

* Fix codeclimate issues

* Fix codeclimate issues

* Fix codeclimate issues

* Fix Travis issue

* Changes in pie plot code

* Remove unnecessary serialization to JSON file

* Move distance and redshift estimations to mchirp_area module

* Fix small errors

* Fix codeclimate issue

* Add color codes for the different CBC sources

* Change option names and help strings

* Assign colors to CBC sources and tag JSON file and plot as EM followup

* Fix codeclimate issues

* Change mc_area_args to be an attribute of LiveEventManager
  • Loading branch information
@veronica-villa
veronica-villa authored and lenona committed Sep 14, 2020
1 parent a9d560f commit 4b7db91
Show file tree
Hide file tree
Showing 4 changed files with 190 additions and 26 deletions.
11 changes: 8 additions & 3 deletions bin/pycbc_live
View file
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@ from pycbc.io.live import SingleCoincForGraceDB
import pycbc.waveform.bank
from pycbc.vetoes.sgchisq import SingleDetSGChisq
from pycbc.waveform.waveform import props
from pycbc import mchirp_area


def ppdets(ifos):
Expand Down Expand Up @@ -63,14 +64,15 @@ def combine_ifar_pvalue(ifar, pvalue, livetime):


class LiveEventManager(object):
def __init__(self, output_path,
def __init__(self, output_path, mc_area_args,
use_date_prefix=False,
ifar_upload_threshold=None,
pval_livetime=None,
enable_gracedb_upload=False,
gracedb_testing=True,
run_snr_optimization=False):
self.path = output_path
self.mc_area_args = mc_area_args

# Figure out what we are supposed to process within the pool of MPI processes
self.comm = mpi.COMM_WORLD
Expand Down Expand Up @@ -241,7 +243,8 @@ class LiveEventManager(object):
event = SingleCoincForGraceDB(live_ifos, coinc_results, bank=bank,
psds=psds, followup_data=fud,
low_frequency_cutoff=f_low,
channel_names=args.channel_name)
channel_names=args.channel_name,
mc_area_args=self.mc_area_args)

end_time = int(coinc_results['foreground/%s/end_time'
% coinc_ifos[0]])
Expand Down Expand Up @@ -606,6 +609,7 @@ LiveSingle.insert_args(parser)
fft.insert_fft_option_group(parser)
Coincer.insert_args(parser)
SingleDetSGChisq.insert_option_group(parser)
mchirp_area.insert_args(parser)
args = parser.parse_args()
scheme.verify_processing_options(args, parser)
fft.verify_fft_options(args, parser)
Expand Down Expand Up @@ -641,7 +645,8 @@ evnt = LiveEventManager(args.output_path,
pval_livetime=args.pvalue_combination_livetime,
enable_gracedb_upload=args.enable_gracedb_upload,
gracedb_testing=not args.enable_production_gracedb_upload,
run_snr_optimization=args.run_snr_optimization)
run_snr_optimization=args.run_snr_optimization,
mc_area_args=mchirp_area.from_cli(args))

sg_chisq = SingleDetSGChisq.from_cli(args, bank, args.chisq_bins)

Expand Down
52 changes: 51 additions & 1 deletion pycbc/io/live.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,6 +3,7 @@
import pycbc
import numpy
import lal
import json
from six import u as unicode
from glue.ligolw import ligolw
from glue.ligolw import lsctables
Expand All @@ -13,7 +14,8 @@
from pycbc import pnutils
from pycbc.tmpltbank import return_empty_sngl
from pycbc.results import ifo_color

from pycbc.results import source_color
from pycbc.mchirp_area import calc_probabilities

#FIXME Legacy build PSD xml helpers, delete me when we move away entirely from
# xml formats
Expand Down Expand Up @@ -104,6 +106,9 @@ def __init__(self, ifos, coinc_results, **kwargs):
channel_names: dict of strings, optional
Strain channel names for each detector.
Will be recorded in the sngl_inspiral table.
mc_area_args: dict of dicts, optional
Dictionary providing arguments to be used in source probability
estimation with pycbc/mchirp_area.py
"""
self.template_id = coinc_results['foreground/%s/template_id' % ifos[0]]
self.coinc_results = coinc_results
Expand Down Expand Up @@ -251,6 +256,17 @@ def __init__(self, ifos, coinc_results, **kwargs):
psds_lal[ifo] = fseries
make_psd_xmldoc(psds_lal, outdoc)

# source probabilities estimation
if 'mc_area_args' in kwargs:
eff_distances = [sngl.eff_distance for sngl in sngl_inspiral_table]
probabilities = calc_probabilities(coinc_inspiral_row.mchirp,
coinc_inspiral_row.snr,
min(eff_distances),
kwargs['mc_area_args'])
self.probabilities = probabilities
else:
self.probabilities = None

self.outdoc = outdoc
self.time = sngl_populated.get_end()

Expand All @@ -265,6 +281,13 @@ def save(self, filename):
gz = filename.endswith('.gz')
ligolw_utils.write_filename(self.outdoc, filename, gz=gz)

# save source probabilities in a json file
if self.probabilities is not None:
prob_fname = filename.replace('.xml.gz', '_probs.json')
with open(prob_fname, 'w') as prob_outfile:
json.dump(self.probabilities, prob_outfile)
logging.info('Source probabilities file saved as %s', prob_fname)

def upload(self, fname, gracedb_server=None, testing=True,
extra_strings=None):
"""Upload this trigger to gracedb
Expand Down Expand Up @@ -332,6 +355,22 @@ def upload(self, fname, gracedb_server=None, testing=True,
pylab.xlabel('Frequency (Hz)')
pylab.ylabel('ASD')
pylab.savefig(psd_series_plot_fname)
pylab.close()

if self.probabilities is not None:
prob_fname = fname.replace('.xml.gz', '_probs.json')
prob_plot_fname = prob_fname.replace('.json', '.png')

prob_plot = {k: v for (k, v) in self.probabilities.items()
if v != 0.0}
labels, sizes = zip(*prob_plot.items())
colors = [source_color(label) for label in labels]
fig, ax = pylab.subplots()
ax.pie(sizes, labels=labels, colors=colors, autopct='%1.1f%%',
textprops={'fontsize': 15})
ax.axis('equal')
fig.savefig(prob_plot_fname)
pylab.close()

gid = None
try:
Expand Down Expand Up @@ -378,6 +417,17 @@ def upload(self, fname, gracedb_server=None, testing=True,
filename=psd_series_plot_fname,
tag_name=['psd'], displayName=['PSDs'])

# upload source probabilities in json format and plot
if self.probabilities is not None:
gracedb.writeLog(gid, 'source probabilities JSON file upload',
filename=prob_fname, tag_name=['em_follow'])
logging.info('Uploaded source probabilities for event %s', gid)
gracedb.writeLog(gid, 'source probabilities plot upload',
filename=prob_plot_fname,
tag_name=['em_follow'])
logging.info('Uploaded source probabilities pie chart for '
'event %s', gid)

except Exception as exc:
logging.error('Something failed during the upload/annotation of '
'event %s on GraceDB. The event may not have been '
Expand Down
139 changes: 117 additions & 22 deletions pycbc/mchirp_area.py
View file
Original file line number Diff line number Diff line change
@@ -1,15 +1,72 @@
# Integration of the area laying in the different cbc regions
# By A. Curiel Barroso
# August 2019
# Module with utilities for estimating candidate events source probabilities
# Initial code by A. Curiel Barroso, August 2019
# Modified by V. Villa-Ortega, January 2020

"""Functions to compute the area corresponding to different CBC on the m1 & m2
plane when given a central mchirp value and uncertainty.
It also includes a function that calculates the source frame when given the
detector frame mass and redshift.
"""

import math
from pycbc.conversions import mass2_from_mchirp_mass1 as m2mcm1
from scipy.integrate import quad
from pycbc.cosmology import _redshift


def insert_args(parser):
mchirp_group = parser.add_argument_group("Arguments for estimating the "
"source probabilities of a "
"candidate event using the snr, "
"mchirp, and effective distance.")
mchirp_group.add_argument('--src-class-mass-range', type=float, nargs=2,
metavar=('MIN_M2', 'MAX_M1'),
default=[1.0, 45.0],
help="Minimum and maximum values for the mass "
"of the binary components, used as limits "
"of the mass plane when computing the area "
"corresponding to different CBC sources.")
mchirp_group.add_argument('--src-class-mass-gap', type=float, nargs=2,
metavar=('MAX_NS', 'MIN_BH'), default=[3.0, 5.0],
help="Limits of the mass gap, that correspond "
"to the maximum mass of a neutron star "
"and the minimum mass of a black hole. "
"Used as limits of integration of the "
"different CBC regions.")
mchirp_group.add_argument('--src-class-mchirp-to-delta', type=float,
metavar='m0', default=0.01,
help='Coefficient to estimate the value of the '
'mchirp uncertainty by mchirp_delta = '
'm0 * mchirp.')
mchirp_group.add_argument('--src-class-eff-to-lum-distance', type=float,
metavar='a0', default=0.759,
help='Coefficient to estimate the value of the '
'luminosity distance from the minimum '
'eff distance by D_lum = a0 * min(D_eff).')
mchirp_group.add_argument('--src-class-lum-distance-to-delta', type=float,
nargs=2, metavar=('b0', 'b1'),
default=[-0.449, -0.342],
help='Coefficients to estimate the value of the '
'uncertainty on the luminosity distance '
'from the estimated luminosity distance and'
' the coinc snr by delta_lum = D_lum * '
'exp(b0) * coinc_snr ** b1.')
mchirp_group.add_argument('--src-class-mass-gap-separate',
action='store_true',
help='Gives separate probabilities for each kind'
' of mass gap CBC sources: GNS, GG, BHG.')


def from_cli(args):
return {'mass_limits': {'max_m1': args.src_class_mass_range[1],
'min_m2': args.src_class_mass_range[0]},
'mass_bdary': {'ns_max': args.src_class_mass_gap[0],
'gap_max': args.src_class_mass_gap[1]},
'estimation_coeff': {'a0': args.src_class_eff_to_lum_distance,
'b0': args.src_class_lum_distance_to_delta[0],
'b1': args.src_class_lum_distance_to_delta[1],
'm0': args.src_class_mchirp_to_delta},
'mass_gap': args.src_class_mass_gap_separate}


def src_mass_from_z_det_mass(z, del_z, mdet, del_mdet):
Expand All @@ -22,23 +79,15 @@ def src_mass_from_z_det_mass(z, del_z, mdet, del_mdet):
return (msrc, del_msrc)


# Integration function
def mchange(x, mc):
"""Returns a component mass as a function of mchirp and the other
component mass.
"""
return m2mcm1(mc, x)


def intmc(mc, x_min, x_max):
"""Returns the integral of mchange between the minimum and maximum values
of a component mass taking mchirp as an argument.
"""Returns the integral of a component mass as a function of the mass of
the other component, taking mchirp as an argument.
"""
integral = quad(mchange, x_min, x_max, args=mc)
integral = quad(lambda x, mc: m2mcm1(mc, x), x_min, x_max, args=mc)
return integral[0]


def calc_areas(trig_mc_det, mass_limits, mass_bdary, z):
def calc_areas(trig_mc_det, mass_limits, mass_bdary, z, mass_gap):
"""Computes the area inside the lines of the second component mass as a
function of the first component mass for the two extreme values
of mchirp: mchirp +/- mchirp_uncertainty, for each region of the source
Expand Down Expand Up @@ -227,12 +276,58 @@ def calc_areas(trig_mc_det, mass_limits, mass_bdary, z):
int_inf_nsbh = ints_nsbh + intline_inf_nsbh

ansbh = int_sup_nsbh - int_inf_nsbh

if mass_gap:
return {
"BNS": abns,
"GNS": agns,
"NSBH": ansbh,
"GG": agg,
"BHG": abhg,
"BBH": abbh
}
return {
"bns": abns,
"gns": agns,
"nsbh": ansbh,
"gg": agg,
"bhg": abhg,
"bbh": abbh
"BNS": abns,
"NSBH": ansbh,
"BBH": abbh,
"Mass Gap": agns + agg + abhg
}


def calc_probabilities(mchirp, snr, eff_distance, src_args):
"""Computes the different probabilities that a candidate event belongs to
each CBC source category taking as arguments the chirp mass, the
coincident SNR and the effective distance, and estimating the
chirp mass uncertainty, the luminosity distance (and its uncertainty)
and the redshift (and its uncertainty). Probability estimation is done
assuming it is directly proportional to the area laying in the
correspondent CBC region.
"""
mass_limits = src_args['mass_limits']
mass_bdary = src_args['mass_bdary']
coeff = src_args['estimation_coeff']
trig_mc_det = {'central': mchirp, 'delta': mchirp * coeff['m0']}
dist_estimation = coeff['a0'] * eff_distance
dist_std_estimation = (dist_estimation * math.exp(coeff['b0']) *
snr ** coeff['b1'])
z_estimation = _redshift(dist_estimation)
z_est_max = _redshift(dist_estimation + dist_std_estimation)
z_est_min = _redshift(dist_estimation - dist_std_estimation)
z_std_estimation = 0.5 * (z_est_max - z_est_min)
z = {'central': z_estimation, 'delta': z_std_estimation}
mass_gap = src_args['mass_gap']

# If the mchirp is greater than the mchirp corresponding to two masses
# equal to the maximum mass, the probability for BBH is 100%
mc_max = mass_limits['max_m1'] / (2 ** 0.2)
if trig_mc_det['central'] > mc_max * (1 + z['central']):
if mass_gap is not False:
probabilities = {"BNS": 0.0, "GNS": 0.0, "NSBH": 0.0, "GG": 0.0,
"BHG": 0.0, "BBH": 1.0}
else:
probabilities = {"BNS": 0.0, "NSBH": 0.0, "BBH": 1.0,
"Mass Gap": 0.0}
else:
areas = calc_areas(trig_mc_det, mass_limits, mass_bdary, z, mass_gap)
total_area = sum(areas.values())
probabilities = {key: areas[key]/total_area for key in areas}
return probabilities
14 changes: 14 additions & 0 deletions pycbc/results/color.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,20 @@
'V1': '#9b59b6', # magenta/purple
}

_source_color_map = {
'BNS': '#A2C8F5', # light blue
'NSBH': '#FFB482', # light orange
'BBH': '#FE9F9B', # light red
'Mass Gap': '#8EE5A1', # light green
'GNS': '#98D6CB', # turquoise
'GG': '#79BB87', # green
'BHG': '#C6C29E' # dark khaki
}


def ifo_color(ifo):
return _ifo_color_map[ifo]


def source_color(source):
return _source_color_map[source]

0 comments on commit 4b7db91

Please sign in to comment.