J/MNRAS/504/3084 ML approach for GRB detection in AstroSat CZTI (Abraham+, 2021)
A machine learning approach for GRB detection in AstroSat CZTI data.
Abraham S., Mukund N., Vibhute A., Sharma V., Iyyani S., Bhattacharya D.,
Rao A.R., Vadawale S. and Bhalerao V.
<Mon. Not. R. Astron. Soc. 504, 3084-3091 (2021)>
=2021MNRAS.504.3084A 2021MNRAS.504.3084A (SIMBAD/NED BibCode)
ADC_Keywords: GRB ; X-ray sources
Keywords: methods: data analysis - methods: statistical - gamma rays: general -
X-rays: bursts
Abstract:
We present a machine learning (ML) based method for automated
detection of Gamma-Ray Burst (GRB) candidate events in the range
60-250 keV from the AstroSat Cadmium Zinc Telluride Imager data. We
use density-based spatial clustering to detect excess power and carry
out an unsupervised hierarchical clustering across all such events to
identify the different light curves present in the data. This
representation helps us to understand the instrument's sensitivity to
the various GRB populations and identify the major non-astrophysical
noise artefacts present in the data. We use Dynamic Time Warping (DTW)
to carry out template matching, which ensures the morphological
similarity of the detected events with known typical GRB light curves.
DTW alleviates the need for a dense template repository often required
in matched filtering like searches. The use of a similarity metric
facilitates outlier detection suitable for capturing previously
unmodelled events. We briefly discuss the characteristics of 35 long
GRB candidates detected using the pipeline and show that with minor
modifications such as adaptive binning, the method is also sensitive
to short GRB events. Augmenting the existing data analysis pipeline
with such ML capabilities alleviates the need for extensive manual
inspection, enabling quicker response to alerts received from other
observatories such as the gravitational-wave detectors.
Description:
AstroSat is India's first multiwavelength space observatory capable
of making observations in X-ray and UV bands. It carries the Cadmium
Zinc Telluride Imager consisting of an array of Cadmium Zinc Telluride
(CZT) detectors, which are pixellated such that each pixel acts as an
independent photon-counting detector.
Thus we construct a template bank for long GRB light curves using 87
known GRBs. The key idea is to minimize the number of templates while
still achieving maximal coverage of the light curves morphologies.
We construct a template bank consisting of 52 GRB light-curve
templates based on the hierarchical clustering analysis results. We
carefully choose these templates to guarantee adequate representation
of all the probable morphologies of GRB events.
The machine learning pipeline detected 223 probable candidates, out of
which 170 were already known to be GRBs. Detailed analysis of the rest
led to the discovery of 35 long GRBs candidate events along with 18
false positives. The table1.dat lists these newly discovered events
along with their trigger time in UTC, T90, peak count rate, total
count rate, mean background counts, and detection significance with
their uncertainties.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
table1.dat 81 35 GRBs candidate events detected with the machine
learning algorithm described in this paper, GCN
circulars have been issued for the highlighted
events
--------------------------------------------------------------------------------
See also:
J/ApJ/888/40 : Fast radio bursts with AstroSat/CZTI (Anumarlapudi+, 2020)
Byte-by-byte Description of file: table1.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 10 A10 --- ID Gamma ray burst GRB ID (GRB ID)
12- 22 A11 "date" Obs.Date UT Date of observation (UTC date)
24- 31 A8 "h:m:s" Obs.Time UT time of exposure start (UTC time)
33- 36 F4.1 s T90 GRB event duration time (T90) (1)
38- 41 F4.2 s e_T90 Mean error on T90 (e_T90)
43- 46 I4 ct/s Peak Count rate peak (Peak)
48- 51 F4.1 ct/s e_Peak Mean error on Peak (e_Peak)
53- 57 I5 ct Total Total of counts (Total counts)
59- 63 F5.1 ct e_Total Mean error on Total (e_Total)
65- 67 I3 ct/s MeanBg Mean of the background count rate
(Mean background) (2)
69- 72 F4.1 ct/s e_MeanBg Mean error of MeanBg (e_MeanBg)
74- 77 F4.1 --- Signi Detection significance (Signi) (3)
79- 81 F3.1 --- e_Signi Mean error (standard deviation) on Signi
(e_Signi) (3)
--------------------------------------------------------------------------------
Note (1): The duration, T90 is calculated as T90 = T95 - T5, where T95 and T5
are the times when 95 percents and 5 percents of the total GRB event
counts are obtained, respectively.
Note (2): The background is modelled by fitting the selected pre- and post-GRB
background regions by a polynomial.
Note (3): The detection significance is calculated as PCR/sqrt(MBR) and the
error on it is obtained by the standard method of error propagation
using the uncertainities reported for PCR and MBR. PCR is the maximum
count rate observed in the light curve of the GRB local. MBR is the
mean background rate obtained by averaging the count rates found in
these regions, subsequently subtracted from the light curve.
--------------------------------------------------------------------------------
History:
From electronic version of the journal
(End) Luc Trabelsi [CDS] 13-May-2024