Machine learning technique for morphological classification of galaxies from the SDSS

Vavilova, I. B.; Dobrycheva, D. V.; Vasylenko, M. Yu.; Elyiv, A.; Мelnyk, О. V.; Khramtsov, Vladislav

doi:10.1051/0004-6361/202038981

Cited by 31 publications

(23 citation statements)

References 88 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…It is clear that the strategy used in this work needs further development to properly assess significantly larger cluster samples or to eventually conduct a blind search for cluster emission. In this respect, machine-based techniques represent an appealing solution to classify the emission in large object samples (e.g., Aniyan & Thorat 2017;Alhassan et al 2018;Domínguez Sánchez et al 2018;Lukic et al 2018Lukic et al , 2019Sadeghi et al 2021;Vavilova et al 2021). As part of this work, we have made public all our images and the detailed results of our decision-tree-based classification, which we hope can provide a good training set for algorithms that attempt either the full classification or to aid the automation at specific intersections in a decision-tree-type approach.…”

Section: Classificationmentioning

confidence: 99%

The Planck clusters in the LOFAR sky

Botteon

Shimwell

Cassano

et al. 2022

A&A

View full text Add to dashboard Cite

Context. Relativistic electrons and magnetic fields permeate the intra-cluster medium (ICM) and manifest themselves as diffuse sources of synchrotron emission observable at radio wavelengths, namely radio halos and radio relics. Although there is broad consensus that the formation of these sources is connected to turbulence and shocks in the ICM, the details of the required particle acceleration, the strength and morphology of the magnetic field in the cluster volume, and the influence of other sources of high-energy particles are poorly known. Aims. Sufficiently large samples of radio halos and relics, which would allow us to examine the variation among the source population and pinpoint their commonalities and differences, are still missing. At present, due to the physical properties of the sources and the capabilities of existing facilities, large numbers of these sources are easiest to detect at low radio frequencies, where they shine brightly. Methods. We examined the low-frequency radio emission from all 309 clusters in the second catalog of Planck Sunyaev Zel’dovich detected sources that lie within the 5634 deg2 covered by the Second Data Release of the LOFAR Two-meter Sky Survey (LoTSS-DR2). We produced LOFAR images at different resolutions, with and without discrete sources subtracted, and created overlays with optical and X-ray images before classifying the diffuse sources in the ICM, guided by a decision tree. Results. Overall, we found 83 clusters that host a radio halo and 26 that host one or more radio relics (including candidates). About half of them are new discoveries. The detection rate of clusters that host a radio halo and one or more relics in our sample is 30 ± 11% and 10 ± 6%, respectively. Extrapolating these numbers, we anticipate that once LoTSS covers the entire northern sky it will provide the detection of 251 ± 92 clusters that host a halo and 83 ± 50 clusters that host at least one relic from Planck clusters alone. All images and results produced in this work are publicly available via the project website.

show abstract

Section: Classificationmentioning

confidence: 99%

The Planck clusters in the LOFAR sky

Botteon

Shimwell

Cassano

et al. 2022

A&A

View full text Add to dashboard Cite

show abstract

“…Techniques based on machine learning may be promising but even then it is apparent that robust classifications are a formidable challenge (see, e.g. Aniyan & Thorat 2017;Lukic et al 2019;Vavilova et al 2021).…”

Section: Source Classificationmentioning

confidence: 99%

Diffuse radio emission from non-Planckgalaxy clusters in the LoTSS-DR2 fields

Hoang

Brüggen

Botteon

et al. 2022

A&A

View full text Add to dashboard Cite

Context. The presence of large-scale magnetic fields and ultra-relativistic electrons in the intra-cluster medium (ICM) is confirmed through the detection of diffuse radio synchrotron sources, so-called radio halos and relics. Due to their steep-spectrum nature, these sources are rarely detected at frequencies above a few gigahertz, especially in low-mass systems. Aims. The aim of this study is to discover and characterise diffuse radio sources in low-mass galaxy clusters in order to understand their origin and their scaling with host cluster properties. Methods. We searched for cluster-scale radio emission from low-mass galaxy clusters in the Low Frequency Array (LOFAR) Twometre Sky Survey -Data Release 2 (LoTSS-DR2) fields. We made use of existing optical (Abell, DESI, WHL) and X-ray (comPRASS, MCXC) catalogues. The LoTSS-DR2 data were processed further to improve the quality of the images that are used to detect and characterise diffuse sources. Results. We detect diffuse radio emission in 28 galaxy clusters. The numbers of confirmed (candidates) halos and relics are six (seven) and 10 (three), respectively. Among these, 11 halos and 10 relics, including candidates, are newly discovered by LOFAR. In addition to these, five diffuse sources are detected in tailed radio galaxies and are probably associated with mergers during the formation of the host clusters. We are unable to classify a further 13 diffuse sources. We compare our newly detected, diffuse sources to known sources by placing them on the scaling relation between the radio power and the mass of the host clusters.

show abstract

“…In recent years in fact, machine learning techniques have seen an increasingly significant impact on astronomical studies, in response to the undergoing rapid growth in size and complexity of datasets as provided by current large surveys like SDSS, MANGA or GAIA (Gaia Collaboration et al 2016), and in preparation for future ones like DESI (Levi et al 2013), SKA (Dewdney et al 2009) and LSST (Ivezic et al 2008). Such algorithms are successfully implemented to solve a variety of different problems, including the classification of galaxy morphological types (e.g., de la Calleja & Fuentes 2004; Barchi et al 2020;Vavilova et al 2021;Reza 2021), the identification of transients (Sooknunan et al 2021), or the multi-parametric analysis of very large databases of galaxy properties (e.g., Teimoorinia et al 2016Teimoorinia et al , 2021Ho 2019;Bluck et al 2019Bluck et al , 2020. Inspired especially by the latter works, in this paper we train and test artificial neural networks (ANN) and random forest decision trees (RF) to assess the performance of a set of carefully selected parameters (both individually and as a whole) and identify which properties are the most relevant in predicting the observed deviation of star-forming galaxies from their average sequence in both the [N ]-and [S ]-BPT diagrams.…”

Section: Introductionmentioning

confidence: 99%

What drives the scatter of local star-forming galaxies in the BPT diagrams? A Machine Learning based analysis

Curti,

Hayden-Pawson,

Maiolino

et al. 2021

Preprint

View full text Add to dashboard Cite

We investigate which physical properties are most predictive of the position of local star forming galaxies on the BPT diagrams, by means of different Machine Learning (ML) algorithms. Exploiting the large statistics from the Sloan Digital Sky Survey (SDSS), we define a framework in which the deviation of star-forming galaxies from their median sequence can be described in terms of the relative variations in a variety of observational parameters. We train artificial neural networks (ANN) and random forest (RF) trees to predict whether galaxies are offset above or below the sequence (via classification), and to estimate the exact magnitude of the offset itself (via regression). We find, with high significance, that parameters associated to variations in the nitrogen-over-oxygen abundance ratio (N/O) are the most predictive for the [N ]-BPT diagram, whereas properties related to star formation (like variations in SFR or EW(Hα)) perform better in the [S ]-BPT diagram. We interpret the former as a reflection of the N/O-O/H relationship for local galaxies, while the latter as primarily tracing the variation in the effective size of the S + emitting region, which directly impacts the [S ] emission lines. This analysis paves the way to assess to what extent the physics shaping local BPT diagrams is also responsible for the offsets seen in high redshift galaxies or, instead, whether a different framework or even different mechanisms need to be invoked.

show abstract

Machine learning technique for morphological classification of galaxies from the SDSS

Cited by 31 publications

References 88 publications

The Planck clusters in the LOFAR sky

The Planck clusters in the LOFAR sky

Diffuse radio emission from non-Planckgalaxy clusters in the LoTSS-DR2 fields

What drives the scatter of local star-forming galaxies in the BPT diagrams? A Machine Learning based analysis

Contact Info

Product

Resources

About