Classification Of Light Curves Research Articles

The ZTF Source Classification Project. I. Methods and Infrastructure

Abstract The Zwicky Transient Facility (ZTF) has been observing the entire northern sky since the start of 2018 down to a magnitude of 20.5 (5σ for 30 s exposure) in the g, r, and i filters. Over the course of two years, ZTF has obtained light curves of more than a billion sources, each with 50–1000 epochs per light curve in g and r, and fewer in i. To be able to use the information contained in the light curves of variable sources for new scientific discoveries, an efficient and flexible framework is needed to classify them. In this paper, we introduce the methods and infrastructure that will be used to classify all ZTF light curves. Our approach aims to be flexible and modular and allows the use of a dynamical classification scheme and labels, continuously evolving training sets, and the use of different machine-learning classifier types and architectures. With this setup, we are able to continuously update and improve the classification of ZTF light curves as new data become available, training samples are updated, and new classes need to be incorporated.

The Automatic Learning for the Rapid Classification of Events (ALeRCE) Alert Broker

Abstract We introduce the Automatic Learning for the Rapid Classification of Events (ALeRCE) broker, an astronomical alert broker designed to provide a rapid and self-consistent classification of large etendue telescope alert streams, such as that provided by the Zwicky Transient Facility (ZTF) and, in the future, the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST). ALeRCE is a Chilean-led broker run by an interdisciplinary team of astronomers and engineers working to become intermediaries between survey and follow-up facilities. ALeRCE uses a pipeline that includes the real-time ingestion, aggregation, cross-matching, machine-learning (ML) classification, and visualization of the ZTF alert stream. We use two classifiers: a stamp-based classifier, designed for rapid classification, and a light curve–based classifier, which uses the multiband flux evolution to achieve a more refined classification. We describe in detail our pipeline, data products, tools, and services, which are made public for the community (see https://alerce.science). Since we began operating our real-time ML classification of the ZTF alert stream in early 2019, we have grown a large community of active users around the globe. We describe our results to date, including the real-time processing of 1.5 × 108 alerts, the stamp classification of 3.4 × 107 objects, the light-curve classification of 1.1 × 106 objects, the report of 6162 supernova candidates, and different experiments using LSST-like alert streams. Finally, we discuss the challenges ahead in going from a single stream of alerts such as ZTF to a multistream ecosystem dominated by LSST.

Exploiting Kepler’s Heritage: A Transfer Learning Approach for Identifying Exoplanets’ Transits in TESS Data

Abstract In the last decade, exoplanets space missions started to collect a huge amount of photometric observations, with over ∼1,000,000 new light curves generated every month from the Transiting Exoplanet Survey Satellite (TESS) full-frame images alone. In order to analyze such an unprecedented volume of data, automated planet-candidate detection has become an appreciable replacement to human vetting. In this work, we present a Machine Learning approach, based on Deep Neural Networks, to perform a binary classification of TESS light curves in terms of planet candidate and not-planet. Since few TESS labeled data exist to date, we pre-train the network with Kepler DR24 data set, including ≳15,000 labeled light curves. Our pre-trained model is then tested on ExoFOP data, showing an appreciable gain in terms of reliability with respect to a randomly initialized model.

On the Classification and Feature Relevance of Multiband Light Curves

Abstract With an expected torrent of data from the Vera C. Rubin Observatory Legacy Survey of Space and Time (LSST), the need for automated identification of noisy and sparse light curves will increase drastically. In this paper, we performed classification of multiband astronomical light curves from the Photometric LSST Astronomical Time-series Classification Challenge (PLAsTiCC) data set via boosted neural nets, boosted decision trees, and a voted classifier for 14 astronomical categories. In order to deal with noisy features, we used wavelet decomposition together with feature selection. We also performed a feature ranking method using a neural network. Our method may be considered an alternative to random forests, which is known to favor features with more categories as relevant. We also investigated the class importance with neural nets using a one-versus-all approach which reduces the multiclass problem to a binary class problem.

Photometric Light Curves Classification with Machine Learning

The Large Synoptic Survey Telescope will begin its survey in 2022 and produce terabytes of imaging data each night. To work with this massive onset of data, automated algorithms to classify astronomical light curves are crucial. Here, we present a method for automated classification of photometric light curves for a range of astronomical objects. Our approach is based on the gradient boosting of decision trees, feature extraction and selection, and augmentation. The solution was developed in the context of The Photometric LSST Astronomical Time Series Classification Challenge (PLAsTiCC) and achieved one of the top results in the challenge.

SuperNNova: an open-source framework for Bayesian, neural network-based supernova classification

ABSTRACTWe introduce SuperNNova, an open-source supernova photometric classification framework that leverages recent advances in deep neural networks. Our core algorithm is a recurrent neural network (RNN) that is trained to classify light curves using only photometric information. Additional information such as host-galaxy redshift can be incorporated to improve performance. We evaluate our framework using realistic supernova simulations that include survey detection. We show that our method, for the type Ia versus non-Ia supernova classification problem, reaches accuracies greater than 96.92 ± 0.09 without any redshift information and up to 99.55 ± 0.06 when redshift, either photometric or spectroscopic, is available. Further, we show that our method attains unprecedented performance for the classification of incomplete light curves, reaching accuracies >86.4 ± 0.1 (>93.5 ± 0.8) without host-galaxy redshift (with redshift information) 2 d before maximum light. In contrast with previous methods, there is no need for time-consuming feature engineering and we show that our method scales to very large data sets with a modest computing budget. In addition, we investigate often neglected pitfalls of machine learning algorithms. We show that commonly used algorithms suffer from poor calibration and overconfidence on out-of-distribution samples when applied to supernova data. We devise extensive tests to estimate the robustness of classifiers and cast the learning procedure under a Bayesian light, demonstrating a much better handling of uncertainties. We study the benefits of Bayesian RNNs for SN Ia cosmology. Our code is open sourced and available on github1.

Supernova Photometric Classification Pipelines Trained on Spectroscopically Classified Supernovae from the Pan-STARRS1 Medium-deep Survey

Abstract Photometric classification of supernovae (SNe) is imperative as recent and upcoming optical time-domain surveys, such as the Large Synoptic Survey Telescope (LSST), overwhelm the available resources for spectrosopic follow-up. Here we develop a range of light curve (LC) classification pipelines, trained on 513 spectroscopically classified SNe from the Pan-STARRS1 Medium-Deep Survey (PS1-MDS): 357 Type Ia, 93 Type II, 25 Type IIn, 21 Type Ibc, and 17 Type I superluminous SNe (SLSNe). We present a new parametric analytical model that can accommodate a broad range of SN LC morphologies, including those with a plateau, and fit this model to data in four PS1 filters (g P1 r P1 i P1 z P1). We test a number of feature extraction methods, data augmentation strategies, and machine-learning algorithms to predict the class of each SN. Our best pipelines result in ≈90% average accuracy, ≈70% average purity, and ≈80% average completeness for all SN classes, with the highest success rates for SNe Ia and SLSNe and the lowest for SNe Ibc. Despite the greater complexity of our classification scheme, the purity of our SN Ia classification, ≈95%, is on par with methods developed specifically for Type Ia versus non-Type Ia binary classification. As the first of its kind, this study serves as a guide to developing and training classification algorithms for a wide range of SN types with a purely empirical training set, particularly one that is similar in its characteristics to the expected LSST main survey strategy. Future work will implement this classification pipeline on ≈3000 PS1/MDS LCs that lack spectroscopic classification.

PELICAN: deeP architecturE for the LIght Curve ANalysis

We developed a deeP architecturE for the LIght Curve ANalysis (PELICAN) for the characterization and the classification of supernovae light curves. It takes light curves as input, without any additional features. PELICAN can deal with the sparsity and the irregular sampling of light curves. It is designed to remove the problem of non-representativeness between the training and test databases coming from the limitations of the spectroscopic follow-up. We applied our methodology on different supernovae light curve databases. First, we tested PELICAN on the Supernova Photometric Classification Challenge for which we obtained the best performance ever achieved with a non-representative training database, by reaching an accuracy of 0.811. Then we tested PELICAN on simulated light curves of the LSST Deep Fields for which PELICAN is able to detect 87.4% of supernovae Ia with a precision higher than 98%, by considering a non-representative training database of 2k light curves. PELICAN can be trained on light curves of LSST Deep Fields to classify light curves of the LSST main survey, which have a lower sampling rate and are more noisy. In this scenario, it reaches an accuracy of 96.5% with a training database of 2k light curves of the Deep Fields. This constitutes a pivotal result as type Ia supernovae candidates from the main survey might then be used to increase the statistics without additional spectroscopic follow-up. Finally we tested PELICAN on real data from the Sloan Digital Sky Survey. PELICAN reaches an accuracy of 86.8% with a training database composed of simulated data and a fraction of 10% of real data. The ability of PELICAN to deal with the different causes of non-representativeness between the training and test databases, and its robustness against survey properties and observational conditions, put it in the forefront of light curve classification tools for the LSST era.

Deep multi-survey classification of variable stars

During the last decade, a considerable amount of effort has been made to classify variable stars using different machine learning techniques. Typically, light curves are represented as vectors of statistical descriptors or features that are used to train various algorithms. These features demand big computational powers that can last from hours to days, making impossible to create scalable and efficient ways of automatically classifying variable stars. Also, light curves from different surveys cannot be integrated and analyzed together when using features, because of observational differences. For example, having variations in cadence and filters, feature distributions become biased and require expensive data-calibration models. The vast amount of data that will be generated soon make necessary to develop scalable machine learning architectures without expensive integration techniques. Convolutional Neural Networks have shown impressing results in raw image classification and representation within the machine learning literature. In this work, we present a novel Deep Learning model for light curve classification, mainly based on convolutional units. Our architecture receives as input the differences between time and magnitude of light curves. It captures the essential classification patterns regardless of cadence and filter. In addition, we introduce a novel data augmentation schema for unevenly sampled time series. We test our method using three different surveys: OGLE-III; Corot; and VVV, which differ in filters, cadence, and area of the sky. We show that besides the benefit of scalability, our model obtains state of the art levels accuracy in light curve classification benchmarks.

From FATS to feets: Further improvements to an astronomical feature extraction tool based on machine learning

Machine learning algorithms are highly useful for the classification of time series data in astronomy in this era of peta-scale public survey data releases. These methods can facilitate the discovery of new unknown events in most astrophysical areas, as well as improving the analysis of samples of known phenomena. Machine learning algorithms use features extracted from collected data as input predictive variables. A public tool called Feature Analysis for Time Series (FATS) has proved an excellent workhorse for feature extraction, particularly light curve classification for variable objects. In this study, we present a major improvement to FATS, which corrects inconvenient design choices, minor details, and documentation for the re-engineering process. This improvement comprises a new Python package called "feets", which is important for future code-refactoring for astronomical software tools.

STACCATO: a novel solution to supernova photometric classification with biased training sets

We present a new solution to the problem of classifying Type Ia supernovae from their light curves alone given a spectroscopically confirmed but biased training set, circumventing the need to obtain an observationally expensive unbiased training set. We use Gaussian processes (GPs) to model the supernovae's (SN) light curves, and demonstrate that the choice of covariance function has only a small influence on the GPs ability to accurately classify SNe. We extend and improve the approach of Richards et al (2012} -- a diffusion map combined with a random forest classifier -- to deal specifically with the case of biassed training sets. We propose a novel method, called STACCATO (SynThetically Augmented Light Curve ClassificATiOn') that synthetically augments a biased training set by generating additional training data from the fitted GPs. Key to the success of the method is the partitioning of the observations into subgroups based on their propensity score of being included in the training set. Using simulated light curve data, we show that STACCATO increases performance, as measured by the area under the Receiver Operating Characteristic curve (AUC), from 0.93 to 0.96, close to the AUC of 0.977 obtained using the 'gold standard' of an unbiased training set and significantly improving on the previous best result of 0.88. STACCATO also increases the true positive rate for SNIa classification by up to a factor of 50 for high-redshift/low brightness SNe.

A Double-line M-dwarf Eclipsing Binary from CSS × SDSS

Abstract Eclipsing binaries offer a unique opportunity to determine basic stellar properties. With the advent of wide-field camera and all-sky time-domain surveys, thousands of eclipsing binaries have been charted via light curve classification, yet their fundamental properties remain unexplored mainly due to the extensive efforts needed for spectroscopic follow-ups. In this paper, we present the discovery of a short-period (P = 0.313 day), double-lined M-dwarf eclipsing binary, CSSJ114804.3+255132/SDSSJ114804.35+255132.6, by cross-matching binary light curves from the Catalina Sky Survey and spectroscopically classified M dwarfs from the Sloan Digital Sky Survey. We obtain follow-up spectra using the Gemini telescope, enabling us to determine the mass, radius, and temperature of the primary and secondary component to be M 1 = 0.47 ± 0.03(statistic) ± 0.03(systematic) M ⊙, M 2 = 0.46 ± 0.03(statistic) ± 0.03(systematic) M ⊙, R 1 = 0.52 ± 0.08(statistic) ± 0.07(systematic) R ⊙, R 2 =0.60 ± 0.08(statistic) ± 0.08(systematic) R ⊙, T 1 = 3560 ± 100 K, and T 2 = 3040 ± 100 K, respectively. The systematic error was estimated using the difference between eccentric and non-eccentric fits. Our analysis also indicates that there is definitively third-light contamination (66%) in the CSS photometry. The secondary star seems inflated, probably due to tidal locking of the close secondary companion, which is common for very short-period binary systems. Future spectroscopic observations with high resolution will narrow down the uncertainties of stellar parameters for both components, rendering this system as a benchmark for studying fundamental properties of M dwarfs.

Double-lined M dwarf eclipsing binaries from Catalina Sky Survey and LAMOST

Eclipsing binaries provide a unique opportunity to determine fundamental stellar properties. In the era of wide-field cameras and all-sky imaging surveys, thousands of eclipsing binaries have been reported through light curve classification, yet their basic properties remain unexplored due to the extensive efforts needed to follow them up spectroscopically. In this paper we investigate three M2-M3 type double-lined eclipsing binaries discovered by cross-matching eclipsing binaries from the Catalina Sky Survey with spectroscopically classified M dwarfs from the Large Sky Area Multi-Object Fiber Spectroscopic Telescope survey data release one and two. Because these three M dwarf binaries are faint, we further acquire radial velocity measurements using GMOS on the Gemini North telescope with 4000, enabling us to determine the mass and radius of individual stellar components. By jointly fitting the light and radial velocity curves of these systems, we derive the mass and radius of the primary and secondary components of these three systems, in the range between 0.28-0.42 and 0.29-0.67, respectively. Future observations with a high resolution spectrograph will help us pin down the uncertainties in their stellar parameters, and render these systems benchmarks to study M dwarfs, providing inputs to improving stellar models in the low mass regime, or establishing an empirical mass-radius relation for M dwarf stars.

CLUSTERING-BASED FEATURE LEARNING ON VARIABLE STARS

ABSTRACT The success of automatic classification of variable stars depends strongly on the lightcurve representation. Usually, lightcurves are represented as a vector of many descriptors designed by astronomers called features. These descriptors are expensive in terms of computing, require substantial research effort to develop, and do not guarantee a good classification. Today, lightcurve representation is not entirely automatic; algorithms must be designed and manually tuned up for every survey. The amounts of data that will be generated in the future mean astronomers must develop scalable and automated analysis pipelines. In this work we present a feature learning algorithm designed for variable objects. Our method works by extracting a large number of lightcurve subsequences from a given set, which are then clustered to find common local patterns in the time series. Representatives of these common patterns are then used to transform lightcurves of a labeled set into a new representation that can be used to train a classifier. The proposed algorithm learns the features from both labeled and unlabeled lightcurves, overcoming the bias using only labeled data. We test our method on data sets from the Massive Compact Halo Object survey and the Optical Gravitational Lensing Experiment; the results show that our classification performance is as good as and in some cases better than the performance achieved using traditional statistical features, while the computational cost is significantly lower. With these promising results, we believe that our method constitutes a significant step toward the automation of the lightcurve classification pipeline.

Modeling lightcurves for improved classification of astronomical objects

Many synoptic surveys are observing large parts of the sky multiple times. The resulting time series of light measurements, called lightcurves, provide a wonderful window to the dynamic nature of the Universe. However, there are many significant challenges in analyzing these lightcurves. We describe a modeling‐based approach using Gaussian process regression for generating critical measures for the classification of such lightcurves. This method has key advantages over other popular nonparametric regression methods in its ability to deal with censoring, a mixture of sparsely and densely sampled curves, the presence of annual gaps caused by objects not being visible throughout the year from a given position on Earth and known but variable measurement errors. We demonstrate that our approach performs better by showing it has a higher correct classification rate than past methods popular in astronomy. Finally, we provide future directions for use in sky‐surveys that are getting even bigger by the day. © 2016 Wiley Periodicals, Inc. Statistical Analysis and Data Mining: The ASA Data Science Journal, 2016

Featureless classification of light curves

In the era of rapidly increasing amounts of time series data, classification of variable objects has become the main objective of time-domain astronomy. Classification of irregularly sampled time series is particularly difficult because the data cannot be represented naturally as a vector which can be directly fed into a classifier. In the literature, various statistical features serve as vector representations. In this work, we represent time series by a density model. The density model captures all the information available, including measurement errors. Hence, we view this model as a generalisation to the static features which directly can be derived, e.g., as moments from the density. Similarity between each pair of time series is quantified by the distance between their respective models. Classification is performed on the obtained distance matrix. In the numerical experiments, we use data from the OGLE and ASAS surveys and demonstrate that the proposed representation performs up to par with the best cur- rently used feature-based approaches. The density representation preserves all static information present in the observational data, in contrast to a less complete description by features. The density representation is an upper boundary in terms of information made available to the classifier. Consequently, the predictive power of the proposed classification depends on the choice of similarity measure and classifier, only. Due to its principled nature, we advocate that this new approach of representing time series has potential in tasks beyond classification, e.g., unsupervised learning.

THE EB FACTORY PROJECT. II. VALIDATION WITH THE KEPLER FIELD IN PREPARATION FOR K2 AND TESS

Large repositories of high precision light curve data, such as the Kepler data set, provide the opportunity to identify astrophysically important eclipsing binary (EB) systems in large quantities. However, the rate of classical "by eye" human analysis restricts complete and efficient mining of EBs from these data using classical techniques. To prepare for mining EBs from the upcoming K2 mission as well as other current missions, we developed an automated end-to-end computational pipeline - the Eclipsing Binary Factory (EBF) - that automatically identifies EBs and classifies them into morphological types. The EBF has been previously tested on ground-based light curves. To assess the performance of the EBF in the context of space-based data, we apply the EBF to the full set of light curves in the Kepler "Q3" Data Release. We compare the EBs identified from this automated approach against the human generated Kepler EB Catalog of ~2,600 EBs. When we require EB classification with at least 90% confidence, we find that the EBF correctly identifies and classifies eclipsing contact (EC), eclipsing semi-detached (ESD), and eclipsing detached (ED) systems with a false positive rate of only 4%, 4%, and 8%, while complete to 64%, 46%, and 32% respectively. When classification confidence is relaxed, the EBF identifies and classifies ECs, ESDs, and EDs with a slightly higher false positive rate of 6%, 16%, and 8%, while much more complete to 86%, 74%, and 62% respectively. Through our processing of the entire Kepler "Q3" dataset, we also identify 68 new candidate EBs that may have been missed by the human generated Kepler EB Catalog. We discuss the EBF's potential application to light curve classification for periodic variable stars more generally for current and upcoming surveys like K2 and the Transiting Exoplanet Survey Satellite.

The VVV Templates Project Towards an automated classification of VVV light-curves

Context. The Vista Variables in the Vía Láctea (VVV) ESO Public Survey is a variability survey of the Milky Way bulge and an adjacent section of the disk carried out from 2010 on ESO Visible and Infrared Survey Telescope for Astronomy (VISTA). The VVV survey will eventually deliver a deep near-IR atlas with photometry and positions in five passbands (ZYJHKS) and a catalogue of 1−10 million variable point sources – mostly unknown – that require classifications.

THE EB FACTORY PROJECT. I. A FAST, NEURAL-NET-BASED, GENERAL PURPOSE LIGHT CURVE CLASSIFIER OPTIMIZED FOR ECLIPSING BINARIES

We describe a new neural-net based light curve classifier and provide it with documentation as a ready-to-use tool for the community. While optimized for identification and classification of eclipsing binary stars, the classifier is general purpose, and has been developed for speed in the context of upcoming massive surveys such as LSST. A challenge for classifiers in the context of neural-net training and massive data sets is to minimize the number of parameters required to describe each light curve. We show that a simple and fast geometric representation that encodes the overall light curve shape, together with a chi-square parameter to capture higher-order morphology information results in efficient yet robust light curve classification, especially for eclipsing binaries. Testing the classifier on the ASAS light curve database, we achieve a retrieval rate of 98\% and a false-positive rate of 2\% for eclipsing binaries. We achieve similarly high retrieval rates for most other periodic variable-star classes, including RR Lyrae, Mira, and delta Scuti. However, the classifier currently has difficulty discriminating between different sub-classes of eclipsing binaries, and suffers a relatively low ($\sim$60\%) retrieval rate for multi-mode delta Cepheid stars. We find that it is imperative to train the classifier's neural network with exemplars that include the full range of light curve quality to which the classifier will be expected to perform; the classifier performs well on noisy light curves only when trained with noisy exemplars. The classifier source code, ancillary programs, a trained neural net, and a guide for use, are provided.

CLASSIFYING SUPERNOVAE USING ONLY GALAXY DATA

We present a new method for probabilistically classifying supernovae (SNe) without using SN spectral or photometric data. Unlike all previous studies to classify SNe without spectra, this technique does not use any SN photometry. Instead, the method relies on host-galaxy data. We build upon the well-known correlations between SN classes and host-galaxy properties, specifically that core-collapse SNe rarely occur in red, luminous, or early-type galaxies. Using the nearly spectroscopically complete Lick Observatory Supernova Search sample of SNe, we determine SN fractions as a function of host-galaxy properties. Using these data as inputs, we construct a Bayesian method for determining the probability that a SN is of a particular class. This method improves a common classification figure of merit by a factor of >2, comparable to the best light-curve classification techniques. Of the galaxy properties examined, morphology provides the most discriminating information. We further validate this method using SN samples from the Sloan Digital Sky Survey and the Palomar Transient Factory. We demonstrate that this method has wide-ranging applications, including separating different subclasses of SNe and determining the probability that a SN is of a particular class before photometry or even spectra can. Since this method uses completely independent data from light-curve techniques, there is potential to further improve the overall purity and completeness of SN samples and to test systematic biases of the light-curve techniques. Further enhancements to the host-galaxy method, including additional host-galaxy properties, combination with light-curve methods, and hybrid methods should further improve the quality of SN samples from past, current, and future transient surveys.