arXiv subjects

P. A. C. Cunha

Publications and source records attributed to P. A. C. Cunha.

17 recordsLinked to original sources

arXiv2026

Euclid Quick Data Release (Q1). Searching for radio-selected \Euclid-dark galaxies in the EDF-N

We present and investigate the properties of a sample of radio-selected, Euclid-dark galaxies, identified from LOFAR HBA observations at 144 MHz within the Euclid Deep Field-North (EDF-N). Starting from radio sources lacking optical counterparts in previous surveys, but detected with Spitzer/IRAC, we identified 166 galaxies with no emission at a more than $3σ$ level in Euclid Quick Release 1 (Q1) images, and no matches in the Euclid Q1 catalogue. To minimise contamination from nearby sources, we selected a sub-sample of 88 isolated galaxies. By exploiting multi-band images and catalogues available for the EDF-N, we inferred the physical properties of our sample via SED fitting. The resulting redshift distribution spans $0.4 \leq z_\mathrm{ph} \leq 5.0$. We used recent sub-arcsecond imaging from the International LOFAR Telescope to constrain the nature of the compact radio emission through brightness temperature estimates. By combining this information with the radio excess relative to the infrared/radio correlation (IRRC), we searched for possible active galactic nuclei (AGN) activity. Approximately 40% of our sources show evidence of AGN activity. The Euclid-dark sources detected in the far-infrared are consistent with a population of heavily obscured, massive star-forming galaxies with high star formation rates. Their location above the star-forming main sequence is consistent with similar near-infrared-dark galaxy populations reported in the literature. We also performed a UV-to-radio median stacking analysis, finding that the two subsamples exhibit similar global physical properties and differ primarily in their radio emission. These preliminary results indicate that the wide area covered by Euclid enables the identification of a higher fraction of systems in which intense star formation and AGN activity coexist, likely capturing a key phase of galaxy--black hole co-evolution.

astro-ph.GA↗

arXiv2026

Euclid Quick Data Release (Q1): Euclid spectroscopy of quasars. 1. Identification and redshift determination of 3500 bright quasars

The slitless spectroscopy mode of the NISP onboard Euclid has enabled efficient spectroscopy of objects within a large FoV. We present a large and homogeneous sample of bright quasars identified from the Euclid Quick Data Release (Q1) by combining high-purity candidate selections from Gaia and WISE with the NISP spectra. Through visual inspection of the Euclid spectra of these quasar candidates, we identify approximately 3500 quasars with reliable redshifts at $0<z\lesssim 4.8$. Of these, 2686 are new spectroscopic identifications relative to existing public compilations. We generate the first Euclid composite spectrum of quasars covering rest-frame NUV to NIR wavelengths without telluric lines, which will be pivotal to NIR quasar spectral analysis. We obtain an empirical spectroscopic depth of $J_{\rm E}\lesssim 21.5$ and $H_{\rm E}\lesssim 21.3$ at the sensitivity of the Wide Field Survey, beyond which the number of securely identified quasars declines sharply. We analyse VIS morphologies using Sersic and CAS metrics, and a deep-learning PSF fraction to track nuclear dominance. At low redshift ($z<0.5$), obvious host structures are common and a single Sersic model fits about half of the sources; at intermediate redshift ($0.5<z<2$), the nuclear component dominates, with 90% of the Sersic fits saturating at the upper index limit. In this intermediate redshift regime, $f_{\rm PSF}$ is available, and we use it as a more reliable compactness measure than the single-Sersic and CAS parameters to quantify nuclear versus host emission. We also explore the novel Euclid NIR colour space and discuss the role of these quasars in refining AGN selection techniques for future Euclid data releases. The spectroscopic bright quasar catalogue of this work, and the composite quasar spectrum, will be available at https://cdsarc.cds.unistra.fr/. (abridged)

astro-ph.GA↗

arXiv2026

The MUSE-Faint survey V. The binary fraction of Leo T

The Leo T dwarf galaxy, the faintest and least massive galaxy known to have recent star formation ($\leq 1~Gyr$), exhibits a high dynamical mass-to-light ratio based on its stellar velocity dispersion ($7.07^{+1.29}_{-1.12}~\mathrm{km\ s^{-1}}$), indicating extreme dark matter dominance. We present the first measurement of the binary fraction of Leo T using MUSE-Faint multi-epoch spectroscopy. We also determine the binary fraction for both young and old stellar populations separately and gain insights into binary properties in more metal-poor environments than the Milky Way or Magellanic Clouds. Finally, we investigate the potential impact of binaries on the inferred stellar velocity dispersion. We employed a forward model methodology combining empirical scaling relations to predict stellar velocity variations and a constrained binary distribution from the literature. To estimate the close binary fraction, we limited the maximum semi-major axis ($a < 10~\mathrm{au}$) and repeated the analysis with a semi-amplitude threshold ($\geq~10~ \mathrm{km\ s^{-1}}$) to check the impact on the inferred stellar velocity dispersion.} The overall binary fraction of Leo T is estimated to be $55^{+40}_{-9} \%$, consistent with similar systems. The close binary fraction ($a < 10~\mathrm{au}$) is $30^{+34}_{-9} \%$, which is aligned with low-metallicity environments. We found a lower binary fraction for the older stellar population ($15^{+43}_{-15} \%$) when compared to the younger population ($35^{+40}_{-6} \%$). Finally, we found no significant inflation of the velocity dispersion estimate due to binary motions when compared to the dispersion inferred from the co-added spectra. This suggests that the co-added spectra effectively provide period-averaged velocities of the stars, thus mitigating the impact of binaries on the overall velocity dispersion measurement.

astro-ph.GA↗

arXiv2025

Euclid Quick Data Release (Q1). Spectroscopic unveiling of highly ionised lines at z = 2.48-3.88

This study explores a rare population of sources in a currently uncharted region of spectroscopic redshift space in the Euclid Quick Data Release (Q1), and is intended potentially to support upcoming spectroscopic studies. Our goal is to identify and investigate a population of sources characterised by highly ionised emission lines in their spectra, which are indicative of active galactic nucleus activity, extreme shock phenomena, or Wolf--Rayet stars. A comprehensive visual inspection of spectra is conducted to ensure the reliability of the sample, focusing on the simultaneous detection of both NeV and OII emission-line measurements, a condition that restricts the Euclid spectroscopic redshift range to z=2.48--3.88. To characterise this population, we analysed the morpho-spectrophotometric properties of their host galaxies. This allowed for a direct comparison with control sources that exhibit similar OII properties and spectroscopic redshifts, but not NeV lines. We identify sources solely based on spectroscopic criteria in the redshift range beyond the Halpha regime. Encompassing 65 potential NeV candidates, the resulting sample delivers the first systematic probe of these NeV candidate emitters at high redshift. We found a good agreement, within 1$σ$, between the spectral measurements calculated using both direct integration and Gaussian fitting methodologies. The NeV candidates exhibit colours similar to bright QSOs, with only a few in the tail of very red quasars. We observed a higher stellar mass content, a lower continuum around the 4000A break, and a similar Sérsic index distribution compared to the control sample. This unique sample paves the way for a wide range of scientific investigations, which will be pursued in the forthcoming data releases.

astro-ph.GA↗

arXiv2025

A gradient boosting and broadband approach to finding Lyman-α emitting galaxies beyond narrowband surveys

In this work, we test whether gradient-boosting algorithms, trained on broadband photometric data from traditional Lyman-$α$ emitting (LAE) surveys, can efficiently and accurately identify LAE candidates from typical star-forming galaxies at similar redshifts and brightness levels. Using galaxy samples at $z \in [2,6]$ derived from the COSMOS2020 and SC4K catalogs, we trained gradient-boosting machine-learning algorithms (LGBM, XGBoost, and CatBoost) using optical and near-infrared broadband photometry. To ensure balanced performance, the models were trained on carefully selected datasets with similar redshift and i-band magnitude distributions. Additionally, the models were tested for robustness by perturbing the photometric data using the associated observational uncertainties. Our classification models achieved F1-scores of $\sim 87\%$ and successfully identified about $7,000$ objects with an unanimous agreement across all models. This more than doubles the number of LAEs identified in the COSMOS field compared with the SC4K dataset. We managed to spectroscopically confirm 60 of these LAE candidates using the publicly available catalogs in the COSMOS field. These results highlight the potential of machine learning in efficiently identifying LAEs candidates. This lays the foundations for applications to larger photometric surveys, such as Euclid and LSST. By complementing traditional approaches and providing robust preselection capabilities, our models facilitate the analysis of these objects. This is crucial to increase our knowledge of the overall LAE population.

astro-ph.GA↗

arXiv2025

Euclid preparation. Estimating galaxy physical properties using CatBoost chained regressors with attention

Euclid will image ~14000 deg^2 of the extragalactic sky at visible and NIR wavelengths, providing a dataset of unprecedented size and richness that will facilitate a multitude of studies into the evolution of galaxies. In the vast majority of cases the main source of information will come from broad-band images and data products thereof. Therefore, there is a pressing need to identify or develop scalable yet reliable methodologies to estimate the redshift and physical properties of galaxies using broad-band photometry from Euclid, optionally including ground-based optical photometry also. To address this need, we present a novel method to estimate the redshift, stellar mass, star-formation rate, specific star-formation rate, E(B-V), and age of galaxies, using mock Euclid and ground-based photometry. The main novelty of our property-estimation pipeline is its use of the CatBoost implementation of gradient-boosted regression-trees, together with chained regression and an intelligent, automatic optimization of the training data. The pipeline also includes a computationally-efficient method to estimate prediction uncertainties, and, in the absence of ground-truth labels, provides accurate predictions for metrics of model performance up to z~2. We apply our pipeline to several datasets consisting of mock Euclid broad-band photometry and mock ground-based ugriz photometry, to evaluate the performance of our methodology for estimating the redshift and physical properties of galaxies detected in the Euclid Wide Survey. The quality of our photometric redshift and physical property estimates are highly competitive overall, validating our modeling approach. We find that the inclusion of ground-based optical photometry significantly improves the quality of the property estimation, highlighting the importance of combining Euclid data with ancillary ground-based optical data. (Abridged)

astro-ph.GA↗

arXiv2025

Euclid Quick Data Release (Q1) Exploring galaxy properties with a multi-modal foundation model

Modern astronomical surveys, such as the Euclid mission, produce high-dimensional, multi-modal data sets that include imaging and spectroscopic information for millions of galaxies. These data serve as an ideal benchmark for large, pre-trained multi-modal models, which can leverage vast amounts of unlabelled data. In this work, we present the first exploration of Euclid data with AstroPT, an autoregressive multi-modal foundation model trained on approximately 300 000 optical and infrared Euclid images and spectral energy distributions (SEDs) from the first Euclid Quick Data Release. We compare self-supervised pre-training with baseline fully supervised training across several tasks: galaxy morphology classification; redshift estimation; similarity searches; and outlier detection. Our results show that: (a) AstroPT embeddings are highly informative, correlating with morphology and effectively isolating outliers; (b) including infrared data helps to isolate stars, but degrades the identification of edge-on galaxies, which are better captured by optical images; (c) simple fine-tuning of these embeddings for photometric redshift and stellar mass estimation outperforms a fully supervised approach, even when using only 1% of the training labels; and (d) incorporating SED data into AstroPT via a straightforward multi-modal token-chaining method improves photo-z predictions, and allow us to identify potentially more interesting anomalies (such as ringed or interacting galaxies) compared to a model pre-trained solely on imaging data.

astro-ph.GA↗

arXiv2025

Euclid Quick Data Release (Q1) First study of red quasars selection

Red quasars constitute an important but elusive phase in the evolution of supermassive black holes, where dust obscuration can significantly alter their observed properties. They have broad emission lines, like other quasars, but their optical continuum emission is significantly reddened, which is why they were traditionally identified based on near- and mid-infrared selection criteria. This work showcases the capability of the \Euclid space telescope to find a large sample of red quasars, using \Euclid near infrared (NIR) photometry. We first conduct a forecast analysis, comparing a synthetic catalogue of red QSOs with COSMOS2020. Using template fitting, we reconstruct \Euclid-like photometry for the COSMOS sources and identify a sample of candidates in a multidimensional colour-colour space achieving $98\%$ completeness for mock red QSOs with $30\%$ contaminants. To refine our selection function, we implement a probabilistic Random Forest classifier, and use UMAP visualisation to disentangle non-linear features in colour-space, reaching $98\%$ completeness and $88\%$ purity. A preliminary analysis of the candidates in the \Euclid Deep Field Fornax (EDF-F) shows that, compared to VISTA+DECAm-based colour selection criteria, \Euclid's superior depth, resolution and optical-to-NIR coverage improves the identification of the reddest, most obscured sources. Notably, the \Euclid exquisite resolution in the $I_E$ filter unveils the presence of a candidate dual quasar system, highlighting the potential for this mission to contribute to future studies on the population of dual AGN. The resulting catalogue of candidates, including more the 150 000 sources, provides a first census of red quasars in \Euclid Q1 and sets the groundwork for future studies in the Euclid Wide Survey (EWS), including spectral follow-up analyses and host morphology characterisation.

astro-ph.GA↗

arXiv2025

Exploring the physical properties of Type II Quasar candidates at intermediate redshifts with CIGALE

Active Galactic Nuclei (AGN) significantly influence galaxy evolution. Specific sources such as obscured AGNs, especially Type II quasars (QSO2), still remain understudied. We characterise 366 QSO2 candidates in the redshift desert (median z~1.1) identified via machine learning from SDSS/WISE photometry, analysing their spectral energy distributions (SEDs) and deriving their physical properties. Using CIGALE, we estimated star formation rate (SFR), stellar mass (M), AGN luminosity, and AGN fraction. We compared these with SPRITZ simulations and the literature, placing results in the galaxy evolution context. Our QSO2 candidates show diverse evolutionary stages. The SFR-M diagram reveals high-SFR sources above the main sequence, linking AGN activity to enhanced star formation. Quenched galaxies may indicate obscured star formation or AGN feedback. Additionally, the physical properties align with SPRITZ composite systems and AGN2, endorsing our obscured AGN classification. This study validates machine learning for identifying AGN-host galaxies, beyond traditional colour-colour selections. Diverse candidate properties highlight this method's ability to identify complex AGN systems. This advances our understanding of AGN-driven galaxy evolution with new target selection.

astro-ph.GA↗

arXiv2024

Euclid preparation. LI. Forecasting the recovery of galaxy physical properties and their relations with template-fitting and machine-learning methods

Euclid will collect an enormous amount of data during the mission's lifetime, observing billions of galaxies in the extragalactic sky. Along with traditional template-fitting methods, numerous machine learning algorithms have been presented for computing their photometric redshifts and physical parameters (PPs), requiring significantly less computing effort while producing equivalent performance measures. However, their performance is limited by the quality and amount of input information, to the point where the recovery of some well-established physical relationships between parameters might not be guaranteed. To forecast the reliability of Euclid photo-$z$s and PPs calculations, we produced two mock catalogs simulating Euclid photometry. We simulated the Euclid Wide Survey (EWS) and Euclid Deep Fields (EDF). We tested the performance of a template-fitting algorithm (Phosphoros) and four ML methods in recovering photo-$z$s, PPs (stellar masses and star formation rates), and the SFMS. To mimic the Euclid processing as closely as possible, the models were trained with Phosphoros-recovered labels. For the EWS, we found that the best results are achieved with a mixed labels approach, training the models with wide survey features and labels from the Phosphoros results on deeper photometry, that is, with the best possible set of labels for a given photometry. This imposes a prior, helping the models to better discern cases in degenerate regions of feature space, that is, when galaxies have similar magnitudes and colors but different redshifts and PPs, with performance metrics even better than those found with Phosphoros. We found no more than 3% performance degradation using a COSMOS-like reference sample or removing u band data, which will not be available until after data release DR1. The best results are obtained for the EDF, with appropriate recovery of photo-$z$, PPs, and the SFMS.

astro-ph.GA↗

arXiv2024

Euclid preparation. XLIX. Selecting active galactic nuclei using observed colours

Euclid will cover over 14000 $deg^{2}$ with two optical and near-infrared spectro-photometric instruments, and is expected to detect around ten million active galactic nuclei (AGN). This unique data set will make a considerable impact on our understanding of galaxy evolution and AGN. In this work we identify the best colour selection criteria for AGN, based only on Euclid photometry or including ancillary photometric observations, such as the data that will be available with the Rubin legacy survey of space and time (LSST) and observations already available from Spitzer/IRAC. The analysis is performed for unobscured AGN, obscured AGN, and composite (AGN and star-forming) objects. We make use of the spectro-photometric realisations of infrared-selected targets at all-z (SPRITZ) to create mock catalogues mimicking both the Euclid Wide Survey (EWS) and the Euclid Deep Survey (EDS). Using these catalogues we estimate the best colour selection, maximising the harmonic mean (F1) of completeness and purity. The selection of unobscured AGN in both Euclid surveys is possible with Euclid photometry alone with F1=0.22-0.23, which can increase to F1=0.43-0.38 if we limit at z>0.7. Such selection is improved once the Rubin/LSST filters (a combination of the u, g, r, or z filters) are considered, reaching F1=0.84 and 0.86 for the EDS and EWS, respectively. The combination of a Euclid colour with the [3.6]-[4.5] colour, which is possible only in the EDS, results in an F1-score of 0.59, improving the results using only Euclid filters, but worse than the selection combining Euclid and LSST. The selection of composite ($f_{\rm AGN}$=0.05-0.65 at 8-40 $μm$) and obscured AGN is challenging, with F1<0.3 even when including ancillary data. This is driven by the similarities between the broad-band spectral energy distribution of these AGN and star-forming galaxies in the wavelength range 0.3-5 $μm$.

astro-ph.GA↗

arXiv2024

Identifying type II quasars at intermediate redshift with few-shot learning photometric classification

We aim to identify QSO2 candidates in the redshift desert using optical and infrared photometry. At this intermediate redshift range, most of the prominent optical emission lines in QSO2 sources (e.g. CIV1549; [OIII]4959,5008) fall either outside the wavelength range of the SDSS optical spectra or in particularly noisy wavelength ranges, making QSO2 identification challenging. Therefore, we adopted a semi-supervised machine learning approach to select candidates in the SDSS galaxy sample. Recent applications of machine learning in astronomy focus on problems involving large data sets, with small data sets often being overlooked. We developed a few-shot learning approach for the identification and classification of rare-object classes using limited training data (200 sources). The new AMELIA pipeline uses a transfer-learning based approach with decision trees, distance-based, and deep learning methods to build a classifier capable of identifying rare objects on the basis of an observational training data set. We validated the performance of AMELIA by addressing the problem of identifying QSO2s at 1 $\leq$ z $\leq$ 2 using SDSS and WISE photometry, obtaining an F1-score above 0.8 in a supervised approach. We then used AMELIA to select new QSO2 candidates in the redshift desert and examined the nature of the candidates using SDSS spectra, when available. In particular, we identified a sub-population of [NeV]3426 emitters at z $\sim$ 1.1, which are highly likely to contain obscured AGNs. We used X-ray and radio cross-matching to validate our classification and investigated the performance of photometric criteria from the literature showing that our candidates have an inherent dusty nature. Finally, we derived physical properties for our QSO2 sample using photoionisation models and verified the AGN classification using an SED fitting.

astro-ph.IM↗

arXiv2023

Selection of powerful radio galaxies with machine learning

We developed and trained a pipeline of three machine learning (ML) models than can predict which sources are more likely to be an AGN and to be detected in specific radio surveys. Also, it can estimate redshift values for predicted radio-detectable AGNs. These models, which combine predictions from tree-based and gradient-boosting algorithms, have been trained with multi-wavelength data from near-infrared-selected sources in the Hobby-Eberly Telescope Dark Energy Experiment (HETDEX) Spring field. Training, testing, calibration, and validation were carried out in the HETDEX field. Further validation was performed on near-infrared-selected sources in the Stripe 82 field. In the HETDEX validation subset, our pipeline recovers 96% of the initially labelled AGNs and, from AGNs candidates, we recover 50% of previously detected radio sources. For Stripe 82, these numbers are 94% and 55%. Compared to random selection, these rates are two and four times better for HETDEX, and 1.2 and 12 times better for Stripe 82. The pipeline can also recover the redshift distribution of these sources with $σ_{\mathrm{NMAD}}$ = 0.07 for HETDEX ($σ_{\mathrm{NMAD}}$ = 0.09 for Stripe 82) and an outlier fraction of 19% (25% for Stripe 82), compatible with previous results based on broad-band photometry. Feature importance analysis stresses the relevance of near- and mid-infrared colours to select AGNs and identify their radio and redshift nature. Combining different algorithms in ML models shows an improvement in the prediction power of our pipeline over a random selection of sources. Tree-based ML models (in contrast to deep learning techniques) facilitate the analysis of the impact that features have on the predictions. This prediction can give insight into the potential physical interplay between the properties of radio AGNs (e.g. mass of black hole and accretion rate).

astro-ph.GA↗

arXiv2022

Euclid preparation: XXII. Selection of Quiescent Galaxies from Mock Photometry using Machine Learning

The Euclid Space Telescope will provide deep imaging at optical and near-infrared wavelengths, along with slitless near-infrared spectroscopy, across ~15,000 sq deg of the sky. Euclid is expected to detect ~12 billion astronomical sources, facilitating new insights into cosmology, galaxy evolution, and various other topics. To optimally exploit the expected very large data set, there is the need to develop appropriate methods and software. Here we present a novel machine-learning based methodology for selection of quiescent galaxies using broad-band Euclid I_E, Y_E, J_E, H_E photometry, in combination with multiwavelength photometry from other surveys. The ARIADNE pipeline uses meta-learning to fuse decision-tree ensembles, nearest-neighbours, and deep-learning methods into a single classifier that yields significantly higher accuracy than any of the individual learning methods separately. The pipeline has `sparsity-awareness', so that missing photometry values are still informative for the classification. Our pipeline derives photometric redshifts for galaxies selected as quiescent, aided by the `pseudo-labelling' semi-supervised method. After application of the outlier filter, our pipeline achieves a normalized mean absolute deviation of ~< 0.03 and a fraction of catastrophic outliers of ~< 0.02 when measured against the COSMOS2015 photometric redshifts. We apply our classification pipeline to mock galaxy photometry catalogues corresponding to three main scenarios: (i) Euclid Deep Survey with ancillary ugriz, WISE, and radio data; (ii) Euclid Wide Survey with ancillary ugriz, WISE, and radio data; (iii) Euclid Wide Survey only. Our classification pipeline outperforms UVJ selection, in addition to the Euclid I_E-Y_E, J_E-H_E and u-I_E,I_E-J_E colour-colour methods, with improvements in completeness and the F1-score of up to a factor of 2. (Abridged)

astro-ph.IM↗

arXiv2022

Improving machine learning-derived photometric redshifts and physical property estimates using unlabelled observations

In the era of huge astronomical surveys, machine learning offers promising solutions for the efficient estimation of galaxy properties. The traditional, `supervised' paradigm for the application of machine learning involves training a model on labelled data, and using this model to predict the labels of previously unlabelled data. The semi-supervised `pseudo-labelling' technique offers an alternative paradigm, allowing the model training algorithm to learn from both labelled data and as-yet unlabelled data. We test the pseudo-labelling method on the problems of estimating redshift, stellar mass, and star formation rate, using COSMOS2015 broad band photometry and one of several publicly available machine learning algorithms, and we obtain significant improvements compared to purely supervised learning. We find that the gradient-boosting tree methods CatBoost, XGBoost, and LightGBM benefit the most, with reductions of up to ~15% in metrics of absolute error. We also find similar improvements in the photometric redshift catastrophic outlier fraction. We argue that the pseudo-labellng technique will be useful for the estimation of redshift and physical properties of galaxies in upcoming large imaging surveys such as Euclid and LSST, which will provide photometric data for billions of sources.

astro-ph.GA↗

arXiv2022

Machine-learning classification of astronomical sources: estimating F1-score in the absence of ground truth

Machine-learning based classifiers have become indispensable in the field of astrophysics, allowing separation of astronomical sources into various classes, with computational efficiency suitable for application to the enormous data volumes that wide-area surveys now typically produce. In the standard supervised classification paradigm, a model is typically trained and validated using data from relatively small areas of sky, before being used to classify sources in other areas of the sky. However, population shifts between the training examples and the sources to be classified can lead to `silent' degradation in model performance, which can be challenging to identify when the ground-truth is not available. In this Letter, we present a novel methodology using the NannyML Confidence-Based Performance Estimation (CBPE) method to predict classifier F1-score in the presence of population shifts, but without ground-truth labels. We apply CBPE to the selection of quasars with decision-tree ensemble models, using broad-band photometry, and show that the F1-scores are predicted remarkably well (MAPE ~ 10%; R^2 = 0.74-0.92). We discuss potential use-cases in the domain of astronomy, including machine-learning model and/or hyperparameter selection, and evaluation of the suitability of training datasets for a particular classification problem.

astro-ph.IM↗

arXiv2022

Photometric redshift-aided classification using ensemble learning

We present SHEEP, a new machine learning approach to the classic problem of astronomical source classification, which combines the outputs from the XGBoost, LightGBM, and CatBoost learning algorithms to create stronger classifiers. A novel step in our pipeline is that prior to performing the classification, SHEEP first estimates photometric redshifts, which are then placed into the data set as an additional feature for classification model training; this results in significant improvements in the subsequent classification performance. SHEEP contains two distinct classification methodologies: (i) Multi-class and (ii) one versus all with correction by a meta-learner. We demonstrate the performance of SHEEP for the classification of stars, galaxies, and quasars using a data set composed of SDSS and WISE photometry of 3.5 million astronomical sources. The resulting F1-scores are as follows: 0.992 for galaxies; 0.967 for quasars; and 0.985 for stars. In terms of the F1-scores for the three classes, SHEEP is found to outperform a recent RandomForest-based classification approach using an essentially identical data set. Our methodology also facilitates model and data set explainability via feature importances; it also allows the selection of sources whose uncertain classifications may make them interesting sources for follow-up observations.

astro-ph.IM↗