Classification Of Variable Stars Research Articles

Machine Learning-based Classification of Variable Stars

The classification of variable stars, essential for revealing information on stellar properties and cosmic distances, traditionally relied on statistical methods and limited data. With the emergence of transformative machine learning methodologies and large stellar surveys, we are able to perform more efficient, accurate, and robust handling of expansive databases. We deploy two different machine learning models—random forest and XGBoost—that effectively classify variable stars identified in the most recent phase of the Optical Gravitational Lensing Experiment using four features: (1) time of minimum brightness, (2) I-band amplitude, (3) mean or maximum I-band magnitude, and (4) period. The two models achieve a cross-correlation score of 99.00%, indicating largely shared classifications. The results produced by the models uncover new insights into the precision of supervised learning models in predicting variable stars with the aid of the most up-to-date survey data.

Automatic Detection of Plateau Phases in Light Curves of Variable Stars

Modern astronomical surveys produce millions of light curves of variable sources. These massive data sets challenge the community to create automatic light-curve processing methods for detection, classification, and characterization of variable stars. In this paper, we present a novel method for extracting the variable components of a light curve based on Otsu’s thresholding method. To validate the effectiveness of this method, we apply it to the light curves of detached eclipsing binaries and dwarf novae, sourced from Optical Gravitational Lensing Experiment catalogs.

Periodic Variable Star Classification with Deep Learning: Handling Data Imbalance in an Ensemble Augmentation Way

Time-domain astronomy is progressing rapidly with the ongoing and upcoming large-scale photometric sky surveys led by the Vera C. Rubin Observatory project (LSST). Billions of variable sources call for better automatic classification algorithms for light curves. Among them, periodic variable stars are frequently studied. Different categories of periodic variable stars have a high degree of class imbalance and pose a challenge to algorithms including deep learning methods. We design two kinds of architectures of neural networks for the classification of periodic variable stars in the Catalina Survey’s Data Release 2: a multi-input recurrent neural network (RNN) and a compound network combing the RNN and the convolutional neural network (CNN). To deal with class imbalance, we apply Gaussian Process to generate synthetic light curves with artificial uncertainties for data augmentation. For better performance, we organize the augmentation and training process in a “bagging-like” ensemble learning scheme. The experimental results show that the better approach is the compound network combing RNN and CNN, which reaches the best result of 86.2% on the overall balanced accuracy and 0.75 on the macro F1 score. We develop the ensemble augmentation method to solve the data imbalance when classifying variable stars and prove the effectiveness of combining different representations of light curves in a single model. The proposed methods would help build better classification algorithms of periodic time series data for future sky surveys (e.g., LSST).

An Evenly Spaced LSST Cadence for Rapidly Variable Stars

Abstract Stars exhibit a bewildering variety of rapidly variable behaviors ranging from explosive magnetic flares to stochastically changing accretion to periodic pulsations or rotation. The principal Rubin Observatory Legacy Survey of Space and Time (LSST) surveys will have cadences too sparse and irregular to capture many of these phenomena. We propose here an LSST microsurvey to observe a single Galactic field, rich in unobscured stars, in a continuous sequence of 30 s exposures for one long winter night in a single photometric band. The result will be a unique data set of ∼1 million regularly spaced stellar light curves. The light curves will constitute a comprehensive collection of late-type stellar flaring, but also other classes like short-period binary systems and cataclysmic variables, young stellar objects, and ultrashort-period exoplanets. An unknown variety of anomalous solar system, Galactic, and extragalactic variables and transients may also be present. A powerful array of statistical procedures can be applied to individual light curves from the long-standing fields of time series analysis, signal processing, and econometrics. Dozens of “features” describing the variability can be extracted and the ensemble of light curves can be subject to advanced machine-learning clustering procedures. This will give a unique, authoritative, objective taxonomy of the rapidly variable sky derived from identically cadenced light curves. This microsurvey is best performed early in the Rubin Observatory program, and the results can inform the wider community on the best approaches to variable star identification and classification from the sparse, irregular cadences that dominate the planned surveys.

A novel approach for variable star classification based on imbalanced learning

Abstract The advent of time-domain sky surveys has generated a vast amount of light variation data, enabling astronomers to investigate variable stars with large-scale samples. However, this also poses new opportunities and challenges for the time-domain research. In this paper, we focus on the classification of variable stars from the Catalina Surveys Data Release 2 and propose an imbalanced learning classifier based on Self-paced Ensemble (SPE) method. Compared with the work of Hosenie et al. (2020), our approach significantly enhances the classification Recall of Blazhko RR Lyrae stars from 12% to 85%, mixed-mode RR Lyrae variables from 29% to 64%, detached binaries from 68% to 97%, and LPV from 87% to 99%. SPE demonstrates a rather good performance on most of the variable classes except RRab, RRc, and contact and semi-detached binary. Moreover, the results suggest that SPE tends to target the minority classes of objects, while Random Forest is more effective in finding the majority classes. To balance the overall classification accuracy, we construct a Voting Classifier that combines the strengths of SPE and Random Forest. The results show that the Voting Classifier can achieve a balanced performance across all classes with minimal loss of accuracy. In summary, the SPE algorithm and Voting Classifier are superior to traditional machine learning methods and can be well applied to classify the periodic variable stars. This paper contributes to the current research on imbalanced learning in astronomy and can also be extended to the time-domain data of other larger sky survey projects (LSST, etc.).

Variable Star Classification with a Multiple-input Neural Network

In this experiment, we created a Multiple-Input Neural Network, consisting of convolutional and multilayer neural networks. With this setup the selected highest-performing neural network was able to distinguish variable stars based on the visual characteristics of their light curves, while taking also into account additional numerical information (e.g., period, reddening-free brightness) to differentiate visually similar light curves. The network was trained and tested on Optical Gravitational Lensing Experiment-III (OGLE-III) data using all OGLE-III observation fields, phase-folded light curves, and period data. The neural network yielded accuracies of 89%–99% for most of the main classes (Cepheids, δ Scutis, eclipsing binaries, RR Lyrae stars, Type-II Cepheids), only the first-overtone anomalous Cepheids had an accuracy of 45%. To counteract the large confusion between the first-overtone anomalous Cepheids and the RRab stars we added the reddening-free brightness as a new input and only stars from the LMC field were retained to have a fixed distance. With this change we improved the neural network’s result for the first-overtone anomalous Cepheids to almost 80%. Overall, the Multiple-input Neural Network method developed by our team is a promising alternative to existing classification methods.

Periodic variable A-F spectral type stars in the northern TESS continuous viewing zone

Context. In the time of large space surveys that provide tremendous amounts of precise data, it is highly desirable to have a commonly accepted methodology and system for the classification of variable stars. This is especially important for A-F stars, which can show intrinsic brightness variations due to both rotation and pulsations. Aims. The goal of our study is to provide a reliable classification of the variability of A-F stars brighter than 11 mag located in the northern TESS continuous viewing zone. We also aim to provide a thorough discussion about issues in the classification related to data characteristics and the issues arising from the similar light-curve shape generated by different physical mechanisms. Methods. We used TESS long- and short-cadence photometric data and corresponding Fourier transform to classify the variability type of the stars. We also used spectroscopic observations to determine the projected rotational velocity of a few stars. Results. We present a clear and concise classification system that is demonstrated on many examples. We find clear signs of variability in 3025 of 5923 studied stars (51%). For 1813 of these 3025 stars, we provide a classification; the rest cannot be unambiguously classified. Of the classified stars, 64.5% are pulsating stars of g-mode γ Doradus (GDOR) and p-mode δ Scuti types and their hybrids. We realised that the long- and short-cadence pre-search data conditioning simple aperture photometry data can differ significantly not only in amplitude but also in the content of instrumental and data-reduction artefacts, making the long-cadence data less reliable. We identified a new group of stars that show stable light curves and characteristic frequency spectrum patterns (8.5% of the classified stars). According to the position in the Hertzsprung-Russell diagram, these stars are likely GDOR stars but are on average about 200 K cooler than GDORs and have smaller amplitudes and longer periods. With the help of spectroscopic measurements of v sin i, we show that the variability of stars with unresolved groups of peaks located close to the positions of the harmonics in their frequency spectra (16% of the classified stars) can be caused by rotation rather than by pulsations. We show that without spectroscopic observations it can be impossible to unambiguously distinguish between ellipsoidal variability and rotational variability. We also applied our methodology to three previous studies and find significant discrepancies in the classification. Conclusions. We demonstrate how difficult the classification of variable A-F stars can be when using only photometric data, how the residual artefacts can produce false positives, and that some types cannot actually be distinguished without spectroscopic observations. Our analysis provides collections that can be used as training samples for automatic classification.

Sparse Identification of Variable Star Dynamics

Variable stars play a crucial role as standard candles and provide valuable insights into stellar physics. They can be modeled either through fully fledged hydrodynamical simulations or analytically as systems of coupled differential equations describing the evolution of relevant physical quantities. Typically, such equations are arrived at by simplified physical assumptions concerning the conservation laws governing stellar interiors. Here we apply a data-driven technique—sparse identification of nonlinear dynamics (SINDy)—to automatically learn governing equations from observed light curves. We apply SINDy to 3100 light curves of three different variable types from the Catalina Sky Survey. The success rate depends systematically on variable type, with possible implications for variable star classification; however, it does not obviously depend on amplitude or period. Successful models can be reduced to the generalized Lienard equation . Members of the Lienard class of ordinary differential equations, such as the well-studied van der Pol oscillator, already saw some application to variable star modeling. For a, b = 0 the equation can be solved exactly, and it admits both periodic and nonperiodic solutions. We find a condition on the coefficients of the general equation for the presence of a limit cycle, which is also observed numerically in several instances.

The Impact of Observing Strategy on the Reliable Classification of Standard Candle Stars: Detection of Amplitude, Period, and Phase Modulation (Blazhko Effect) of RR Lyrae Stars with LSST

Abstract The Vera C. Rubin Observatory will carry out its Legacy Survey of Space and Time (LSST) with a single-exposure depth of r ∼ 24.7 and an anticipated baseline of 10 yr, allowing access to the Milky Way’s old halo not only deeper than, but also with a longer baseline and better cadence than, e.g., PS1 3π. This will make the LSST ideal to study populations of variable stars such as RR Lyrae stars (RRL). Here, we address the question of observing strategy optimization of LSST, as survey footprint definition, single-visit exposure time, as well as the cadence of repeat visits in different filters are yet to be finalized. We present metrics used to assess the impact of different observing strategies on the reliable detectability and classification of standard candle variable stars, including detection of amplitude, period, and phase modulation effects of RRL (the so-called Blazhko effect), by evaluating metrics for simulated potential survey designs. So far, due to the depths and cadences of typical all-sky surveys, it has been nearly impossible to study this effect on a larger sample. All-sky surveys with relatively few observations over a moderately long baseline allow only for fitting phase-folded RRL light curves, thus integrating over the complete survey length and hiding any information regarding possible period or phase modulation during the survey. On the other hand, surveys with cadences fit to detect slightly changing light curves usually have a relatively small footprint. LSST’s survey strategy, however, will allow for studying variable stars in a way that makes population studies possible.

Classification of Variable Stars Light Curves Using Long Short Term Memory Network

Owing to the current and upcoming extensive surveys studying the stellar variability, accurate and quicker methods are required for the astronomers to automate the classification of variable stars. The traditional approach of classification requires the calculation of the period of the observed light curve and assigning different variability patterns of phase folded light curves to different classes. However, applying these methods becomes difficult if the light curves are sparse or contain temporal gaps. Also, period finding algorithms start slowing down and become redundant in such scenarios. In this work, we present a new automated method, 1D CNN-LSTM, for classifying variable stars using a hybrid neural network of one-dimensional CNN and LSTM network which employs the raw time-series data from the variable stars. We apply the network to classify the time-series data obtained from the OGLE and the CRTS survey. We report the best average accuracy of 85% and F1 score of 0.71 for classifying five classes from the OGLE survey. We simultaneously apply other existing classification methods to our dataset and compare the results.

Optical and Near-infrared Pulsation Properties of RR Lyrae and Population II Cepheid Variables in the Messier 15 Globular Cluster

Abstract Messier 15 (NGC 7078) is an old and metal-poor post core-collapse globular cluster that hosts a rich population of variable stars. We report new optical (gi) and near-infrared (NIR, JK s ) multi-epoch observations for 129 RR Lyrae, 4 Population II Cepheids (3 BL Herculis, 1 W Virginis), and 1 anomalous Cepheid variable candidate in M15 obtained using the MegaCam and the WIRCam instruments on the 3.6 m Canada–France–Hawaii Telescope. Multi-band data are used to improve the periods and classification of variable stars, and determine accurate mean magnitudes and pulsational amplitudes from the light curves fitted with optical and NIR templates. We derive optical and NIR period–luminosity relations for RR Lyrae stars which are best constrained in the K s band, m K s = − 2.333 ( 0.054 ) log P + 13.948 ( 0.015 ) with a scatter of only 0.037 mag. Theoretical and empirical calibrations of RR Lyrae period–luminosity–metallicity relations are used to derive a true distance modulus to M15: 15.196 ± 0.026 (statistical) ± 0.039 (systematic) mag. Our precise distance moduli based on RR Lyrae stars and Population II Cepheid variables are mutually consistent and agree with recent distance measurements in the literature based on Gaia parallaxes and other independent methods.

Variable star classification across the Galactic bulge and disc with the VISTA Variables in the Vía Láctea survey

Abstract We present VIVACE, the VIrac VAriable Classification Ensemble, a catalogue of variable stars extracted from an automated classification pipeline for the Vista Variables in the Vía Láctea (VVV) infrared survey of the Galactic bar/bulge and southern disc. Our procedure utilises a two-stage hierarchical classifier to first isolate likely variable sources using simple variability summary statistics and training sets of non-variable sources from the Gaia early third data release, and then classify candidate variables using more detailed light curve statistics and training labels primarily from OGLE and VSX. The methodology is applied to point-spread-function photometry for ∼490 million light curves from the VIRAC v2 astrometric and photometric catalogue resulting in a catalogue of ∼1.4 million likely variable stars, of which ∼39, 000 are high-confidence (classification probability >0.9) RR Lyrae ab stars, ∼8000 RR Lyrae c/d stars, ∼187, 000 detached/semi-detached eclipsing binaries, ∼18, 000 contact eclipsing binaries, ∼1400 classical Cepheid variables and ∼2200 Type II Cepheid variables. Comparison with OGLE-4 suggests a completeness of around 90 per cent for RRab and ≲ 60 per cent for RRc/d, and a misclassification rate for known RR Lyrae stars of around 1 per cent for the high confidence sample. We close with two science demonstrations of our new VIVACE catalogue: first, a brief investigation of the spatial and kinematic properties of the RR Lyrae stars within the disc/bulge, demonstrating the spatial elongation of bar-bulge RR Lyrae stars is in the same sense as the more metal-rich red giant population whilst having a slower rotation rate of ∼40 km s−1kpc−1; and secondly, an investigation of the Gaia EDR3 parallax zeropoint using contact eclipsing binaries across the Galactic disc plane and bulge.

Classification of periodic variable stars with novel cyclic-permutation invariant neural networks

ABSTRACT We present Cyclic-Permutation Invariant Neural Networks, a novel class of neural networks (NNs) designed to be invariant to phase shifts of period-folded periodic sequences by means of ‘symmetry padding’. In the context of periodic variable star light curves, initial phases are exogenous to the physical origin of the variability and should thus be immaterial to the downstream inference application. Although previous work utilizing NNs commonly operated on period-folded light curves, no approach to date has taken advantage of such a symmetry. Across three different data sets of variable star light curves, we show that two implementations of Cyclic-Permutation Invariant Networks—iTCN and iResNet—consistently outperform state-of-the-art non-invariant baselines and reduce overall error rates by between 4 to 22 per cent. Over a 10-class OGLE-III sample, the iTCN/iResNet achieves an average per-class accuracy of 93.4 per cent/93.3 per cent, compared to recurrent NN/random forest accuracies of 70.5 per cent/89.5 per cent in a recent study using the same data. Finding improvement on a non-astronomy benchmark, we suggest that the methodology introduced here should also be applicable to a wide range of science domains where periodic data abounds.

Informative Bayesian model selection for RR Lyrae star classifiers

ABSTRACT Machine learning has achieved an important role in the automatic classification of variable stars, and several classifiers have been proposed over the last decade. These classifiers have achieved impressive performance in several astronomical catalogues. However, some scientific articles have also shown that the training data therein contain multiple sources of bias. Hence, the performance of those classifiers on objects not belonging to the training data is uncertain, potentially resulting in the selection of incorrect models. Besides, it gives rise to the deployment of misleading classifiers. An example of the latter is the creation of open-source labelled catalogues with biased predictions. In this paper, we develop a method based on an informative marginal likelihood to evaluate variable star classifiers. We collect deterministic rules that are based on physical descriptors of RR Lyrae stars, and then, to mitigate the biases, we introduce those rules into the marginal likelihood estimation. We perform experiments with a set of Bayesian logistic regressions, which are trained to classify RR Lyraes, and we found that our method outperforms traditional non-informative cross-validation strategies, even when penalized models are assessed. Our methodology provides a more rigorous alternative to assess machine learning models using astronomical knowledge. From this approach, applications to other classes of variable stars and algorithmic improvements can be developed.

Automatic catalog of RR Lyrae from ∼14 million VVV light curves: How far can we go with traditional machine-learning?

Context. The creation of a 3D map of the bulge using RR Lyrae (RRL) is one of the main goals of the VISTA Variables in the Via Lactea Survey (VVV) and VVV(X) surveys. The overwhelming number of sources undergoing analysis undoubtedly requires the use of automatic procedures. In this context, previous studies have introduced the use of machine learning (ML) methods for the task of variable star classification. Aims. Our goal is to develop and test an entirely automatic ML-based procedure for the identification of RRLs in the VVV Survey. This automatic procedure is meant to be used to generate reliable catalogs integrated over several tiles in the survey. Methods. Following the reconstruction of light curves, we extracted a set of period- and intensity-based features, which were already defined in previous works. Also, for the first time, we put a new subset of useful color features to use. We discuss in considerable detail all the appropriate steps needed to define our fully automatic pipeline, namely: the selection of quality measurements; sampling procedures; classifier setup, and model selection. Results. As a result, we were able to construct an ensemble classifier with an average recall of 0.48 and average precision of 0.86 over 15 tiles. We also made all our processed datasets available and we published a catalog of candidate RRLs. Conclusions. Perhaps most interestingly, from a classification perspective based on photometric broad-band data, our results indicate that color is an informative feature type of the RRL objective class that should always be considered in automatic classification methods via ML. We also argue that recall and precision in both tables and curves are high-quality metrics with regard to this highly imbalanced problem. Furthermore, we show for our VVV data-set that to have good estimates, it is important to use the original distribution more abundantly than reduced samples with an artificial balance. Finally, we show that the use of ensemble classifiers helps resolve the crucial model selection step and that most errors in the identification of RRLs are related to low-quality observations of some sources or to the increased difficulty in resolving the RRL-C type given the data.

Image-based Classification of Variable Stars: First Results from Optical Gravitational Lensing Experiment Data

Abstract Recently, machine learning methods have presented a viable solution for the automated classification of image-based data in various research fields and business applications. Scientists require a fast and reliable solution in order to handle increasingly large amounts of astronomical data. However, so far astronomers have been mainly classifying variable starlight curves based on various pre-computed statistics and light curve parameters. In this work we use an image-based Convolutional Neural Network to classify the different types of variable stars. We use images of phase-folded light curves from the Optical Gravitational Lensing Experiment (OGLE)-III survey for training, validating, and testing, and use OGLE-IV survey as an independent data set for testing. After the training phase, our neural network was able to classify the different types between 80% and 99%, and 77%–98%, accuracy for OGLE-III and OGLE-IV, respectively.

Imbalance learning for variable star classification

ABSTRACT The accurate automated classification of variable stars into their respective subtypes is difficult. Machine learning–based solutions often fall foul of the imbalanced learning problem, which causes poor generalization performance in practice, especially on rare variable star subtypes. In previous work, we attempted to overcome such deficiencies via the development of a hierarchical machine learning classifier. This ‘algorithm-level’ approach to tackling imbalance yielded promising results on Catalina Real-Time Survey (CRTS) data, outperforming the binary and multiclass classification schemes previously applied in this area. In this work, we attempt to further improve hierarchical classification performance by applying ‘data-level’ approaches to directly augment the training data so that they better describe underrepresented classes. We apply and report results for three data augmentation methods in particular: Randomly Augmented Sampled Light curves from magnitude Error (RASLE), augmenting light curves with Gaussian Process modelling (GpFit) and the Synthetic Minority Oversampling Technique (SMOTE). When combining the ‘algorithm-level’ (i.e. the hierarchical scheme) together with the ‘data-level’ approach, we further improve variable star classification accuracy by 1–4 per cent. We found that a higher classification rate is obtained when using GpFit in the hierarchical model. Further improvement of the metric scores requires a better standard set of correctly identified variable stars, and perhaps enhanced features are needed.

Scalable end-to-end recurrent neural network for variable star classification

ABSTRACT During the last decade, considerable effort has been made to perform automatic classification of variable stars using machine-learning techniques. Traditionally, light curves are represented as a vector of descriptors or features used as input for many algorithms. Some features are computationally expensive, cannot be updated quickly and hence for large data sets such as the LSST cannot be applied. Previous work has been done to develop alternative unsupervised feature extraction algorithms for light curves, but the cost of doing so still remains high. In this work, we propose an end-to-end algorithm that automatically learns the representation of light curves that allows an accurate automatic classification. We study a series of deep learning architectures based on recurrent neural networks and test them in automated classification scenarios. Our method uses minimal data pre-processing, can be updated with a low computational cost for new observations and light curves, and can scale up to massive data sets. We transform each light curve into an input matrix representation whose elements are the differences in time and magnitude, and the outputs are classification probabilities. We test our method in three surveys: OGLE-III, Gaia, and WISE. We obtain accuracies of about $95{{\ \rm per\ cent}}$ in the main classes and $75{{\ \rm per\ cent}}$ in the majority of subclasses. We compare our results with the Random Forest classifier and obtain competitive accuracies while being faster and scalable. The analysis shows that the computational complexity of our approach grows up linearly with the light-curve size, while the traditional approach cost grows as Nlog (N).

Streaming classification of variable stars

ABSTRACTIn the last years, automatic classification of variable stars has received substantial attention. Using machine learning techniques for this task has proven to be quite useful. Typically, machine learning classifiers used for this task require to have a fixed training set, and the training process is performed offline. Upcoming surveys such as the Large Synoptic Survey Telescope will generate new observations daily, where an automatic classification system able to create alerts online will be mandatory. A system with those characteristics must be able to update itself incrementally. Unfortunately, after training, most machine learning classifiers do not support the inclusion of new observations in light curves, they need to re-train from scratch. Naively re-training from scratch is not an option in streaming settings, mainly because of the expensive pre-processing routines required to obtain a vector representation of light curves (features) each time we include new observations. In this work, we propose a streaming probabilistic classification model; it uses a set of newly designed features that work incrementally. With this model, we can have a machine learning classifier that updates itself in real time with new observations. To test our approach, we simulate a streaming scenario with light curves from Convention, Rotation and planetary Transits (CoRoT), Orbital Gravitational Lensing Experiment (OGLE), and Massive Compact Halo Object (MACHO) catalogues. Results show that our model achieves high classification performance, staying an order of magnitude faster than traditional classification approaches.

Calibrating long-period variables as standard candles with machine learning

Variable stars with well-calibrated period-luminosity relationships provide accurate distance measurements to nearby galaxies and are therefore a vital tool for cosmology and astrophysics. While these measurements typically rely on samples of Cepheid and RR-Lyrae stars, abundant populations of luminous variable stars with longer periods of $10 - 1000$ days remain largely unused. We apply machine learning to derive a mapping between lightcurve features of these variable stars and their magnitude to extend the traditional period-luminosity (PL) relation commonly used for Cepheid samples. Using photometric data for long period variable stars in the Large Magellanic cloud (LMC), we demonstrate that our predictions produce residual errors comparable to those obtained on the corresponding Cepheid population. We show that our model generalizes well to other samples by performing a blind test on photometric data from the Small Magellanic Cloud (SMC). Our predictions on the SMC again show small residual errors and biases, comparable to results that employ PL relations fitted on Cepheid samples. The residual biases are complementary between the long period variable and Cepheid fits, which provides exciting prospects to better control sources of systematic error in cosmological distance measurements. We finally show that the proposed methodology can be used to optimize samples of variable stars as standard candles independent of any prior variable star classification.