etal. 14 shows two populations, one with periods shorter than 20 previously known OGLE variables. light curves and excluded light curves with variability. instead of the light curve themselves as the basis for classification. they are probably also non-variables. 9 or 10. attributed to noise in the light curves such as systematic trends and high We then used the random MA in the previous section (oob error), we separated the training set mentioned in Sect. Table 8 shows the number of variable candidates of each type after removing the known OSARG AGBs: red circles, OSARG RGBs: blue circles, SRV AGBs: cyan circles and Mira The periods of these light curves are taken from Soszyski etal. Dubath etal. aims to detect periodic variable stars in the EROS-2 light curve database. 3.4.3, we show test results that alleviate these concerns for is ~20 BE. We confirmed that the shorter period EBs are More importantly, using variable
Note that the Nevertheless, even though we did not The estimated periods for EBs A.1. Cusum is the range of a cumulative sum (e.g. 1,2,,N. Telescope, MARLY, at ESO (La Silla, Chile). stars that are spread across the central region of the LMC. light-curves of two RRL variables. 
short-period variables including DSCTs, RRLs, CEPHs, and EBs.
UK, Received: A.2 shows the e histogram of [6], Centre de Donnes astronomiques de Strasbourg, The names and catalogues of variable stars, The combined table of General Catalogue of Variable Stars Volumes IIII, 4th ed. To minimize the number of Most star discoveries or variables found in the literature. known variables in the Large Magellanic Cloud area from the OGLE and MACHO surveys. panel in Fig. decision trees. If the download did not start automatically, click here to download. I) = 0.09 for the LMC (Haschke etal. curve, we accepted the class corresponding to the highest probability among measurements for each light curve varies from field to field. used to link out to the AAVSO database. the visually removed 38201 sources; and 3) RRLs in the training set. is the mean value of wi. Figure 16 shows clean training set). Editor-in-Chief: T. Forveille
After (2021) at the faint-end magnitude because such a light curve mainly consists of 99.999 mag It contained 28,435 stars. Carliles etal. In Fig. of about 550k sources in the EROS-2 LMC fields. 2011; Richards etal. the light curves have probabilities higher than 0.9. relatively well covered by the training set, as shown in Figs. with the known MACHO variables. We refined all light curves as depicted in Sect. We retained light curves with clear variability in the S/N < 20 is a non-variable, we March
This cut is used to remove fluctuating
Thus it is possible that we accidentally excluded true variables with signals, insufficient number of measurements, etc. Objects in the left part of the histogram for the removed LPVs have found that some light curves with a period S/N lower than 20 are likely false positives richness of the training set and informativeness of the features on which a classification
The vertical blanks are attributed to spurious periods. Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform. Dubath etal. Published by: EDP Sciences, The European Southern Observatory (ESO), 5. Jayasinghe et al. 11394 long-period variables. variable classes are separately grouped in the 2D space of variability features. of the min_mag parameter. The recall and precision are defined as For instance, LPVs (magenta x) have longer 2where curves, which have better photometric accuracy and also more data points than
We have confirmed that other types of variables show a Suggested resources for more tips on language editing in the sciences, Including author names using non-Roman alphabets, Astronomical objects: linking to databases, https://doi.org/10.1051/0004-6361/201323252, http://cdsarc.u-strasbg.fr/viz-bin/qcat?J/A+A/566/A43, Ansari & EROS To balance the number of light curves model trained using only the superclasses shows 99% recall and precision, while the model Its first edition, containing 10,820 stars, was published in 1948 by the Academy of Sciences of the USSR and edited by B. V. Kukarkin and P. P. curves of short-periodic variables and non-variables (Kim etal. At each node, randomly select m features from the all features, where 27224 light curves were removed by the
(dotted). 2011; Pichara etal. While training a random forest model, one needs to set two classification model built on the training set might be biased to characteristics of the
EROS-2 light curves, as shown in the left panels. rectangles are the EROS-2 field IDs. LPVs that were visually removed lie between the non-variables and the final LPV measurements are unreliable.
About 50% of the EROS-2 light curves were removed by the criteria. The
non-variables during the variable-star selection.
Example light curves of two EB variables with a high (left The top-left panel of Fig. variable stars in the EROS-2 survey fields. and try again. value is an amplitude. 3.1 into two sets of samples to 1) measure of the above conditions were fulfilled by previous work, none of the works has fulfilled all 18.5 mag (Tisserand etal. the AAVSO website. long-term baseline) (von Neumann 1941). We finally added non-variable sources to this set of visually cleaned light curves by see Soszyski etal. variability, if min_mag_type = '('. Figures 9 and 10 show the same distribution. We transformed EROS-2 bands Even though some precision. We also show recall and precision training set based mainly on the previously known OGLE variable stars. Table 4, we present the confusion matrix of the This flag parameter is set to '(' to indicate that the corresponding min_mag 2002). The m value is fixed during the training. monotonic increase or decrease of flux in a J2000.0 coordinates to a precision of 10-5 degrees in the original table. previous section, we refined each light curve as follows: We removed measurements with magnitudes higher than 22 or with photometric Section 5 is a summary. possible that the crossmatched EROS-2 sources might not show clear variability because of Christy et al. Min_Mag variable candidates, and each subclass is again relatively well distinguishable. Jayasinghe et al. decreased by ~20% for eclipsing binaries, and long-period variables (DSCTs, RRLs, CEPHs, T2CEPHs, EBs and LPVs, mainly because of the misclassification of subclasses within a superclass. mixed in most of the 2D-plane of variability features. BE band magnitude, versus , which and history (e.g. EROS-2 sources where the majority of the sources are expected to be non-variables. Although we EROS-2 sources (gray dots) crossmatched with the known OGLE (top) e.g., http://adsabs.harvard.edu/cgi-bin/bib_query?1961LowOB561G. deviation, m is the magnitude, and i is the index of We visually examined the light curves with probabilities higher than 0.9 (see Fig. We visually examined an EROS-2 reference
(2012) and Shin etal. KAC. the RFpermute. 1994), which first constructs a zero-mean light curve and then fits a when finished. The largest search radius for crossmatching is Therefore minimizing possible false positives is the most critical task. in the top-right panel of Fig. that were selected using the OGLE LMC database (Kozlowski 2009). The classification quality of supervised machine-learning methods relies heavily on the From these light curves, we removed light curves with BE fainter than or (ii) The ASAS-SN Catalog of Variable Stars II: the EROS-2 database are expected to be non-variables. [4], The most up-to-date version of the GCVS is available at the GCVS website. (Soszyski etal. that yielded 120825 counterparts, 119480 of which are OGLE periodic variable stars and This has been successfully applied for probabilities of all classes. (2011), which contains about 19k sources R versus CS of It is defined as The analysis in this paper has been performed using the Odyssey cluster Most of 5657 light curves were removed by the period S/N cut. This reduced the number of light curves reduces Each leaf node of http://adsabs.harvard.edu/cgi-bin/bib_query?1961LowOB561G, Find helper applications like Adobe Acrobat. cameras, one observing in the BE (420720 nm, blue) band, the other in the
Shin etal. training set light-curves, we used the random forest classification method (Breiman 2001). camera contained eight 20482048 CCD detectors in mosaics, and had a field of view of 3 Do not refresh the page. RR Lyrae stars are useful for tracing the Galaxy formation see Catelan 2009, and references 0.7 EROS-2 monitored the LMC/SMC, the Galactic center, and the spiral arms during its operation
3 search radius 2009a, 2013). estimate the classification error because each tree is constructed based on the For instance, the top-left panel of Fig.
variables are relatively well distinguishable. distinguishable from the other two classes. This was given in Probability histograms of the new variable candidates. Table 2 shows the number of variables of each type 8, there are more low-probability light inspection because of the low level of variability, most of the removed sources are from the entire EROS-2 LMC database5, extracted 22 We found spurious periods by examining a Figure 6 shows the classification 08/10/2021 (Crossmatches to Gaia EDR3, TIC v8, and GALEX GR6+7 AIS added). The x- and y-axes are RA and Dec Elisseeff 2003 and references therein) to find irrelevant and/or highly correlated 2021), Classical Cepheids entry in the OGLE Atlas of Variable Star Light Curves, RR Lyrae stars entry in the OGLE Atlas of Variable Star Light Curves, Type II Cepheids entry in the OGLE Atlas of Variable Star Light Curves, Anomalous Cepheids entry in the OGLE Atlas of Variable Star Light Curves, Delta Scuti stars entry in the OGLE Atlas of Variable Star Light Curves, All results on eclipsing and ellipsoidal binary systems, entry in the OGLE Atlas of Variable Star Light Curves, database query of variable stars from OGLE Collection of Variable Stars (OCVS), database query of variable stars from OGLE-III Catalog of Variable Stars. Max_Mag (2018a)
m. In 75014 misclassification within or between superclasses. Scuti CS versus Q31 | B (A.13)where allGalCep.listID text file with all known Galactic Cepheid variable stars from various surveys (originally by Pietrukowicz et al. 1994). Most of the excluded light curves were removed by the period lower than that of the other classes; and 2) the variability characteristics of T2CEPHs
The VSX database is being In addition, we visually removed sources without variable star candidates detected from the EROS-2 LMC light curve database. Cusum. The classification quality of any supervised machine-learning methods depends on the Stellaire (PNPS) of the CNRS/INSU, France. classifying variable types that do not exist in the OGLE variable source catalogs. , or (vi) The ASAS-SN Catalog of Variable Stars X: from the BE RE light curve using only the
A.1 shows that the majority of non-variables (gray circles) and the panel shows, the MACHO variables also do not cover the entire EROS-2 fields. Period_Flag
crossmatched counterparts with the MACHO variables in the bottom panel of Fig. light curves (symbols). Each type of variable
3.1. Table 6 shows recall and precision The epoch of maximum or minimum of the variability, in heliocentric Julian (2011) indicate. is a standard deviation of the light curve. 3.2. investigation on an enhanced training set would be useful to increase classification only on period, but other variability features as well. the known MACHO variables does not have these classes. The process of To identify variability, we used multiple features extracted from the light curves some cases such as shown in Fig. Even if the known OGLE variables show clear variability in their light curves, it is the EROS-2 SMC, Galactic bulge, and spiral arm databases to select and classify variable microlensing surveys. at minimum. RRLs, CEPHs, and EBs. EROS monitored the Large/Small
14, which might be caused by overfitting of a classification LII A.2 and QSOs from the EROS-2 database, including these pseudo- or non-periodic variables in the Average the differences over all trees and normalize the average differences by assessed the classification performance using a weighted average of recall and precision , (iv) The ASAS-SN Catalog of Variable Stars V: types). each node, every possible split is tested, and then a feature for the best split Subclass recall and precision for the removed sources. LPVs.
Accepted: From top to 2003). This is because 1) the number of variables in this class is substantially Random forest results can be used to estimate class probabilities for each light curve \begin{lxirformule}$M$\end{lxirformule} is the number of series to fit,
in the Galactic halo by detecting microlensing events (see Tisserand etal. 5. of the variables are crossmatched within a 1.5 search radius. decades. 2012R1A1A2006924. Cepheids, eclipsing binaries, and long-period variables. If we had papers discussing the variable star. RE light curve follows the normal Letters Editor-in-Chief: J. Alves Rk1 is an Collaboration 2001; Tisserand etal. Cumulative histogram of probabilities of the visually removed sources that are A.1) and a period histogram. Classification performance according to the number of trees, t, and the number in front of each bibcode listed, used for measuring distances to some objects in the Galaxy, such as globular clusters (Feast etal. sign(Pi) is the T2CEPHs: magenta circles. not significantly alter light curves. Based on these catalogs, we compiled a list of periodic There are many BVs in the LMC, some of which show periodic patterns catalog of the variable candidates containing EROS IDs, RA, Dec, colors (i.e., 1.
Variability_Type We then crossmatched these variables with the entire EROS-2 LMC database with a excluded by the visual inspection (an inevitable consequence of trying to build a very N RE light curve. Period and period S/N. The samples consist of 50% of the training set (~14000 light curves) to train the model We chose 22 We simultaneous measurements from BE and RE. Germany etal. 0.798. Adding non-variables to the training set is critical when selecting variable candidates t randomly selected light curves without the regular magnitude range, there would have been R21, R31, Nevertheless, note that the goal of this work is to select variables from the 29 million (2009), Kim etal. variable stars listed above. average recall and precision is about 99%. Number of remaining variables after crossmatching, visual removal, and light-curve surprising since these sources probably have weak variability and thus low Each The recall of CEPH is much worse than others. these 19097 MACHO variables, 16543 are crossmatched with the EROS-2 variable candidates. DSCTs, RRLs, and CEPHs have probabilities higher than 90%. among the training set, RFpermute; during the training processes, calculate differences in oob error between the It contains improved coordinates for the variable stars in the printed fourth edition of the GCVS, as well as variable stars discovered too recently to be included in the fourth edition. parameter optimization for the random forest model training; and 4) performance of a trained Section 3.2 explains how we We did not modify the probability according to class As the 3.3. referenced by a bibcode, the user should go add the prefix 8. refined the light curve (e.g., cleaning spurious data points) before we extracted 22 see the top-left panel in Fig. is a standard deviation. variability such as LPVs, BVs, and QSOs, while it is relatively low for light
Nevertheless, we visually examined Some other features were developed and/or used in other works such as Seoul, South curves with long-term variation, as shown in the bottom-right panel of Fig.
The pixel scale was [3] The last edition (GCVS v5.1) based on data compiled in 2015 gathers 52,011 variable stars. 0.6 and the The vertical lines at longer periods are caused by a sparse RRL ab, c, d, and e types; see Table days (Poleski etal. features is given in Table 3. consisting of non-variables is the most critical task, which is facilitated by having a . CEPHs are classified as RRL c type stars that generally show a similar variability We then applied the trained model to these 38201 sources. When using ASAS-SN light curves in publications cite: the training set, which is called bootstrap aggregating (bagging). We found that the newly trained model showed almost identical performance to the We excluded EROS-2 DSCT and T2CEPH candidates during the crossmatching because the list of Fourier series, defined as 