J/MNRAS/493/1926 4FGL blazar classification neural network (Kovacevic+, 2020)
Classification of blazar candidates of uncertain type from the Fermi LAT 8-yr
source catalogue with an artificial neural network.
Kovacevic M., Chiaro G., Cutini S., Tosti G.
<Mon. Not. R. Astron. Soc., 493, 1926-1935 (2020)>
=2020MNRAS.493.1926K 2020MNRAS.493.1926K (SIMBAD/NED BibCode)
ADC_Keywords: Active gal. nuclei ; BL Lac objects ; Gamma rays
Keywords: methods: statistical - galaxies: active -
BL Lacertae objects: general - gamma rays: galaxies
Abstract:
The Fermi Large Area Telescope (LAT) has detected more than 5000
gamma-ray sources in its first 8 years of operation. More than 3000 of
them are blazars. About 60 per cent of the Fermi-LAT blazars are
classified as BL Lacertae objects (BL Lacs) or Flat Spectrum Radio
Quasars (FSRQs), while the rest remain of uncertain type. The goal of
this study was to classify those blazars of uncertain type, using a
supervised machine learning method based on an artificial neural
network, by comparing their properties to those of known gamma-ray
sources. Probabilities for each of 1329 uncertain blazars to be a BL
Lac or FSRQ are obtained. Using 90 per cent precision metric, 801 can
be classified as BL Lacs and 406 as FSRQs while 122 still remain
unclassified. This approach is of interest because it gives a fast
preliminary classification of uncertain blazars. We also explored how
different selections of training and testing samples affect the
classification and discuss the meaning of network outputs.
Description:
In this paper we use artificial neural network (ANN), a supervised
machine learning method, in order to classify blazars of uncertain
type (BCUs) in the Fermi LAT 8-yr source catalog (4FGL), into BL Lacs
and FSRQs. The 4FGL catalog contains various parameters obtained from
LAT observation of the whole sky, mainly in the 100MeV-100GeV
band. We use fluxes in different energy bands (spectra) and fluxes in
different annual time bins (light curves, variability) as input
parameters for the ANN algorithm.
The final result is the probability for each of 1329 BCU to be BL Lac
or FSRQ. Ratio of BL Lac to FSRQ candidates is about 1:1 for the
Galactic plane and about 2:1 for the whole sky.
File Summary:
--------------------------------------------------------------------------------
FileName Lrecl Records Explanations
--------------------------------------------------------------------------------
ReadMe 80 . This file
table2.dat 51 1329 *BL Lac probability for each BCU in 4FGL
catalog (v19) in ascending order
--------------------------------------------------------------------------------
Note on table2.dat: FSRQ probability is not in the table. It is obtained by
subtracting BL Lac probability from number 1.
--------------------------------------------------------------------------------
See also:
J/ApJ/810/14 : Third catalog of LAT-detected AGNs (3LAC) (Ackermann+, 2015)
J/ApJS/218/23 : Fermi LAT third source catalog (3FGL) (Acero+, 2015)
J/ApJ/820/8 : 3FGL sources statistical classifications
(Saz Parkinson+, 2016)
J/MNRAS/462/3180 : 3FGL Blazar of Unknown Type classification (Chiaro+, 2016)
Byte-by-byte Description of file: table2.dat
--------------------------------------------------------------------------------
Bytes Format Units Label Explanations
--------------------------------------------------------------------------------
1- 17 A17 --- Name BCU name in 4FGL catalog (4FGL_JHHMM.m+DDMM)
19- 25 F7.3 deg GLAT Galactic latitude (1)
27- 33 F7.3 deg GLON Galactic longitude (1)
35- 39 F5.3 --- Lb [0/1]=+ BL Lac probability (2)
41- 45 F5.3 --- b_Lb [0/1] Lower limit of error interval of Lb (3)
47- 51 F5.3 --- B_Lb [0/1] Upper limit of error interval of Lb (3)
--------------------------------------------------------------------------------
Note (1): GLAT and GLON are the ones reported in 4FGL catalog.
Note (2): Lb for each BCU is average value of 300 different Lb, each
corresponding to different choice for train and test sample.
Note (3): Error interval of Lb for each BCU is 1 sigma interval of 300
different Lb.
--------------------------------------------------------------------------------
Acknowledgements:
Milos Kovacevic, milos.kovacevic(at)pg.infn.it
(End) Milos Kovacevic [INFN Perugia], Patricia Vannier [CDS] 14-Apr-2020