We have investigated a number of factors that can have significant impacts on the classification performance of $γ$-ray sources detected by Fermi Large Area Telescope (LAT) with machine learning techniques. We show that a framework of automatic feature selection can construct a simple model with a small set of features which yields better performance over previous results. Secondly, because of the small sample size of the training/test sets of certain classes in $γ$-ray, nested re-sampling and cross-validations are suggested for quantifying the statistical fluctuations of the quoted accuracy. We have also constructed a test set by cross-matching the identified active galactic nuclei (AGNs) and the pulsars (PSRs) in the Fermi LAT eight-year point source catalog (4FGL) with those unidentified sources in the previous 3$^{\rm rd}$ Fermi LAT Source Catalog (3FGL). Using this cross-matched set, we show that some features used for building classification model with the identified source can suffer from the problem of covariate shift, which can be a result of various observational effects. This can possibly hamper the actual performance when one applies such model in classifying unidentified sources. Using our framework, both AGN/PSR and young pulsar (YNG)/millisecond pulsar (MSP) classifiers are automatically updated with the new features and the enlarged training samples in 4FGL catalog incorporated. Using a two-layer model with these updated classifiers, we have selected 20 promising MSP candidates with confidence scores $>98\%$ from the unidentified sources in 4FGL catalog which can provide inputs for a multi-wavelength identification campaign.
Machine learning, algorithms to extract empirical knowledge from data, can be used to classify data, which is one of the most common tasks in observational astronomy. In this paper, we focus on Bayesian data classification algorithms using the Gaussian mixture model and show two applications in pulsar astronomy. After reviewing the Gaussian mixture model and the related Expectation-Maximization algorithm, we present a data classification method using the Neyman-Pearson test. To demonstrate the method, we apply the algorithm to two classification problems. Firstly, it is applied to the well kno…
Data collection for scientific applications is increasing exponentially and is forecasted to soon reach peta- and exabyte scales. Applications which process and analyze scientific data must be scalable and focus on execution performance to keep pace. In the field of radio astronomy, in addition to increasingly large datasets, tasks such as the identification of transient radio signals from extrasolar sources are computationally expensive. We present a scalable approach to radio pulsar detection written in Scala that parallelizes candidate identification to take advantage of in-memory task proc…
There are many occasions when one does not have complete information in order to classify objects into different classes, and yet it is important to do the best one can since other decisions depend on that. In astronomy, especially time-domain astronomy, this situation is common when a transient is detected and one wishes to determine what it is in order to decide if one must follow it. We propose to use the Difference Boosting Neural Network (DBNN) which can boost differences between feature vectors of different objects in order to differentiate between them. We apply it to the publicly avail…
PESIT South Campus（PESIT南校区）；University of Calcutta（加尔各答大学）
We present the results of various automated classification methods, based on machine learning (ML), of objects from data releases 6 and 7 (DR6 and DR7) of the Sloan Digital Sky Survey (SDSS), primarily distinguishing stars from quasars. We provide a careful scrutiny of approaches available in the literature and have highlighted the pitfalls in those approaches based on the nature of data used for the study. The aim is to investigate the appropriateness of the application of certain ML methods. The manuscript argues convincingly in favor of the efficacy of asymmetric AdaBoost to classify photom…
We describe the application of data mining algorithms to research problems in astronomy. We posit that data mining has always been fundamental to astronomical research, since data mining is the basis of evidence-based discovery, including classification, clustering, and novelty discovery. These algorithms represent a major set of computational tools for discovery in large databases, which will be increasingly essential in the era of data-intensive astronomy. Historical examples of data mining in astronomy are reviewed, followed by a discussion of one of the largest data-producing projects anti…
In the last decade a new generation of telescopes and sensors has allowed the production of a very large amount of data and astronomy has become a data-rich science. New automatic methods largely based on machine learning are needed to cope with such data tsunami. We present some results in the fields of photometric redshifts and galaxy classification, obtained using the MLPQNA algorithm available in the DAMEWARE (Data Mining and Web Application Resource) for the SDSS galaxies (DR9 and DR10). We present PhotoRApToR (Photometric Research Application To Redshift): a Java based desktop applicatio…
The advent of synoptic sky surveys has spurred the development of techniques for real-time classification of astronomical sources in order to ensure timely follow-up with appropriate instruments. Previous work has focused on algorithm selection or improved light curve representations, and naively convert light curves into structured feature sets without regard for the time span or phase of the light curves. In this paper, we highlight the violation of a fundamental machine learning assumption that occurs when archival light curves with long observational time spans are used to train classifier…
Self-supervised learning for radio-astronomy source classification: a benchmark
University of Catania（卡塔尼亚大学）；INAF - Osservatorio Astrofisico di Catania（意大利国家天体物理研究所卡塔尼亚天文台）；Macquarie University（麦考瑞大学）；University of Milano Bicocca（米兰比可卡大学）
In this work we introduce group-equivariant self-attention models to address the problem of explainable radio galaxy classification in astronomy. We evaluate various orders of both cyclic and dihedral equivariance, and show that including equivariance as a prior both reduces the number of epochs required to fit the data and results in improved performance. We highlight the benefits of equivariance when using self-attention as an explainable model and illustrate how equivariant models statistically attend the same features in their classifications as human astronomers.
We introduce a new generic Archetype technique for source classification and identification, based on the NP-complete set cover problem (SCP) in computer science and operations research (OR). We have developed a new heuristic SCP solver, by combining the greedy algorithm and the Lagrangian Relaxation (LR) approximation method. We test the performance of our code on the test cases from Beasley's OR Library and show that our SCP solver can efficiently yield solutions that are on average 99% optimal in terms of the cost. We discuss how to adopt SCP for classification purposes and put forward a ne…
In recent years, Imaging Atmospheric Cherenkov Telescopes (IACTs) have discovered a rich diversity of very high energy (VHE, > 100 GeV) gamma-ray emitters in the sky. These instruments image Cherenkov light emitted by gamma-ray induced particle cascades in the atmosphere. Background from the much more numerous cosmic-ray cascades is efficiently reduced by considering the shape of the shower images, and the capability to reduce this background is one of the key aspects that determine the sensitivity of a IACT. In this work we apply a tree classification method to data from the High Energy Stere…
PESIT South Campus（PESIT南校区）；Jain University（贾因大学）
Astroinformatics is a new impact area in the world of astronomy, occasionally called the final frontier, where several astrophysicists, statisticians and computer scientists work together to tackle various data intensive astronomical problems. Exponential growth in the data volume and increased complexity of the data augments difficult questions to the existing challenges. Classical problems in Astronomy are compounded by accumulation of astronomical volume of complex data, rendering the task of classification and interpretation incredibly laborious. The presence of noise in the data makes ana…
This article is based on the tutorial we gave at the hands-on workshop of the ICRANet-ISFAHAN Astronomy Meeting. We first introduce the basic theory of machine learning and sort out the whole process of training a neural network. We then demonstrate this process with an example of inferring redshifts from SDSS spectra. To emphasize that machine learning for astronomy is easy to get started, we demonstrate that the most basic CNN network can be used to obtain high accuracy, we also show that with simple modifications, the network can be converted for classification problems and also to processi…
Machine learning (automated processes that learn by example in order to classify, predict, discover or generate new data) and artificial intelligence (methods by which a computer makes decisions or discoveries that would usually require human intelligence) are now firmly established in astronomy. Every week, new applications of machine learning and artificial intelligence are added to a growing corpus of work. Random forests, support vector machines, and neural networks (artificial, deep, and convolutional) are now having a genuine impact for applications as diverse as discovering extrasolar p…
Where do cosmic X-rays come from? Every new unidentified X-ray source could potentially revolutionize our understanding of the universe. The international collaborative astronomy project EXTraS aimed at automatically classifying new sources of X-ray emission (e.g., stars or galaxies) in the large observation database of the X-ray satellite XMM-Newton. Because data archives have reached dimensions of big data astronomers used different machine-learning (ML) random forest decision tree algorithms that performed the classification process. In this bachelor thesis in information design, I was inte…
Advances on the classification of radio image cubes
Bernoulli Institute for Mathematics, Computer Science and Artificial Intelligence（伯努利数学、计算机科学与人工智能研究所）；University of Groningen（格罗宁根大学）；University of Stellenbosch（斯泰伦博斯大学）；ASTRON（荷兰射电天文研究所）
High-volume feature-rich data sets are becoming the bread-and-butter of 21st century astronomy but present significant challenges to scientific discovery. In particular, identifying scientifically significant relationships between sets of parameters is non-trivial. Similar problems in biological and geosciences have led to the development of systems which can explore large parameter spaces and identify potentially interesting sets of associations. In this paper, we describe the application of automated discovery systems of relationships to astronomical data sets, focussing on an evolutionary p…
Light Curve Classification with DistClassiPy: a new distance-based classifier
University of Delaware（特拉华大学）；Indian Institute of Science Education and Research, Bhopal（印度科学教育与研究学院博帕尔分校）；California Institute of Technology（加州理工学院）；Inter University Centre for Astronomy and Astrophysics (IUCAA)（大学间天文与天体物理中心）；Vera C. Rubin Observatory（维拉·C·鲁宾天文台）
针对时域天文学大数据的分类需求，开发了基于距离度量的开源分类器 DistClassiPy，在 6,000 颗变星上使用 18 种度量实现了达 SOTA 性能、低计算成本和高可解释性的光变曲线分类。
THESEUS is a medium size space mission of the European Space Agency, currently under evaluation for a possible launch in 2032. Its main objectives are to investigate the early Universe through the observation of gamma-ray bursts and to study the gravitational waves electromagnetic counterparts and neutrino events. On the other hand, its instruments, which include a wide field of view X-ray (0.3-5 keV) telescope based on lobster-eye focusing optics and a gamma-ray spectrometer with imaging capabilities in the 2-150 keV range, are also ideal for carrying out unprecedented studies in time domain …
Astronomy has entered the multi-messenger data era and Machine Learning has found widespread use in a large variety of applications. The exploitation of synoptic (multi-band and multi-epoch) surveys, like LSST (Legacy Survey of Space and Time), requires an extensive use of automatic methods for data processing and interpretation. With data volumes in the petabyte domain, the discrimination of time-critical information has already exceeded the capabilities of human operators and crowds of scientists have extreme difficulty to manage such amounts of data in multi-dimensional domains. This work i…
Neutrino Telescope Event Classification on Quantum Computers
University of the Basque Country UPV/EHU（巴斯克大学）；TECNALIA, Basque Research and Technology Alliance (BRTA)（德克拉尼亚，巴斯克研究与技术联盟）；Harvard University（哈佛大学）；IKERBASQUE, Basque Foundation for Science（伊克尔巴斯基，巴斯克科学基金会）；Basque Center for Applied Mathematics (BCAM)（巴斯克应用数学中心）
We apply a combination of a Genetic Algorithms (GA) and Support Vector Machines (SVM) machine learning algorithm to solve two important problems faced by the astronomical community: star/galaxy separation, and photometric redshift estimation of galaxies in survey catalogs. We use the GA to select the relevant features in the first step, followed by optimization of SVM parameters in the second step to obtain an optimal set of parameters to classify or regress, in process of which we avoid over-fitting. We apply our method to star/galaxy separation in Pan-STARRS1 data. We show that our method co…
Evaluating Bayesian deep learning for radio galaxy classification
Jodrell Bank Centre for Astrophysics（焦德雷尔班克天体物理中心）；Department of Physics & Astronomy（物理与天文学院）；University of Manchester（曼彻斯特大学）；The Alan Turing Institute（阿兰·图灵研究所）
Exploring the Universe with SNAD: Anomaly Detection in Astronomy
Space Research Institute of the Russian Academy of Sciences（俄罗斯科学院空间研究所）；University of Illinois at Urbana-Champaign（伊利诺伊大学厄巴纳-香槟分校）；Center for AstroPhysical Surveys（天体物理巡天中心）；National Center for Supercomputing Applications（国家超级计算应用中心）；Lomonosov Moscow State University（罗蒙诺索夫莫斯科国立大学）；Université Clermont Auvergne（克莱蒙奥弗涅大学）；CNRS/IN2P3（法国国家科学研究中心/核物理与粒子物理研究所）；Sternberg Astronomical Institute（斯特恩伯格天文研究所）；National Research University Higher School of Economics（国立研究型大学高等经济学院）
本文综述国际项目 SNAD 如何结合主动学习等机器学习方法，在大规模天文巡天中检测异常，推动天文现象发现分类及机器学习在天体物理中的应用。
Automated spectral classification is an active research area in astronomy at the age of data explosion. While new generation of sky survey telescopes (e.g. LAMOST and SDSS) produce huge amount of spectra, automated spectral classification is highly required to replace the current model fitting approach with human intervention. Galaxies, and especially active galactic nucleus (AGNs), are important targets of sky survey programs. Efficient and automated methods for galaxy spectra classification is the basis of systematic study on physical properties and evolution of galaxies. To address the prob…
The changing heavens have played a central role in the scientific effort of astronomers for centuries. Galileo's synoptic observations of the moons of Jupiter and the phases of Venus starting in 1610, provided strong refutation of Ptolemaic cosmology. In more modern times, the discovery of a relationship between period and luminosity in some pulsational variable stars led to the inference of the size of the Milky Way, the distance scale to the nearest galaxies, and the expansion of the Universe. Distant explosions of supernovae were used to uncover the existence of dark energy and provide a pr…
We describe in general terms the practical use in astronomy of a higher-order statistical quantity called Spectral Kurtosis (SK), and describe the first implementation of SK-enabled firmware in the F-engine (Fourier transform-engine) of a digital FX correlator for Expanded Owens Valley Solar Array (EOVSA). The development of the theory for SK is summarized, leading to an expression for generalized SK that is applicable to both SK spectrometers and those not specifically designed for SK. We also give the means for computing both the SK estimator and thresholds for its application as a discrimin…
We present an application of a particular machine-learning method (Boosted Decision Trees, BDTs using AdaBoost) to separate stars and galaxies in photometric images using their catalog characteristics. BDTs are a well established machine learning technique used for classification purposes. They have been widely used specially in the field of particle and astroparticle physics, and we use them here in an optical astronomy application. This algorithm is able to improve from simple thresholding cuts on standard separation variables that may be affected by local effects such as blending, badly cal…
This project outlines the complete development of a variable star classification algorithm methodology. With the advent of Big-Data in astronomy, professional astronomers are left with the problem of how to manage large amounts of data, and how this deluge of information can be studied in order to improve our understanding of the universe. While our focus will be on the development of machine learning methodologies for the identification of variable star type based on light curve data and associated information, one of the goals of this work is the acknowledgment that the development of a true…
Constructing Impactful Machine Learning Research for Astronomy: Best Practices for Researchers and Reviewers
SRON Netherlands Institute for Space Research（SRON荷兰空间研究所）；Anton Pannekoek Institute for Astronomy, University of Amsterdam（阿姆斯特丹大学安东·潘涅库克天文研究所）；Space Telescope Science Institute（空间望远镜科学研究所）；Johns Hopkins University（约翰斯·霍普金斯大学）；Sorbonne Université（索邦大学）；Max Plank Institute for Astronomy（马克斯·普朗克天文研究所）；Fermi National Accelerator Laboratory（费米国家加速器实验室）；Kavli Institute for Cosmological Physics, University of Chicago（芝加哥大学卡夫利宇宙物理研究所）；University of Chicago（芝加哥大学）；University of Manchester（曼彻斯特大学）
We present new fully-automatic classification model to select extragalactic objects within astronomy photometric catalogs. Construction of the our classification model is based on the three important procedures: 1) data representation to create feature space; 2) building hypersurface in feature space to limit range of features (outliers detection); 3) building hyperplane separating extragalactic objects from the galactic ones. We trained our model with 1.7 million objects (1.4 million galaxies and quasars, 0.3 million stars). The application of the model is presented as a photometric catalog o…
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 genera…
Istituto Nazionale di Fisica Nucleare（意大利国家核物理研究所）
The detection of gravitational waves has inaugurated the era of gravitational astronomy and opened new avenues for the multimessenger study of cosmic sources. Thanks to their sensitivity, the Advanced LIGO and Advanced Virgo interferometers will probe a much larger volume of space and expand the capability of discovering new gravitational wave emitters. The characterization of these detectors is a primary task in order to recognize the main sources of noise and optimize the sensitivity of interferometers. Glitches are transient noise events that can impact the data quality of the interferomete…
NASA Goddard Space Flight Center（美国国家航空航天局戈达德太空飞行中心）；Cavendish Laboratory（卡文迪许实验室）；Kavli Institute for Cosmology（卡弗里宇宙学研究所）
We present the first public release of our generic neural network training algorithm, called SkyNet. This efficient and robust machine learning tool is able to train large and deep feed-forward neural networks, including autoencoders, for use in a wide range of supervised and unsupervised learning applications, such as regression, classification, density estimation, clustering and dimensionality reduction. SkyNet uses a `pre-training' method to obtain a set of network parameters that has empirically been shown to be close to a good solution, followed by further optimisation using a regularised…
