Inferring physical properties of galaxies from their emission-line spectra

Abstract

We present a new approach based on Supervised Machine Learning (SML) algorithms to infer key physical properties of galaxies (density, metallicity, column density and ionization parameter) from their emission line spectra. We introduce a numerical code (called game, GAlaxy Machine learning for Emission lines) implementing this method and test it extensively. game delivers excellent predictive performances, especially for estimates of metallicity and column densities. We compare game with the most widely used diagnostics (e.g. R 23 , [NII]λ6584 / Hα indicators) showing that it provides much better accuracy and wider applicability range. game is particularly suitable for use in combination with Integral Field Unit (IFU) spectroscopy, both for restframe optical/UV nebular lines and far-infrared/sub-mm lines arising from PhotoDissociation Regions. Finally, game can also be applied to the analysis of synthetic galaxy maps built from numerical simulations.

Search for Galaxy Cluster Candidates in the Cosmic Microwave Background Maps of the Planck Space Mission Using a Convolutional Neural Network Based on the Method of Tracing the Sunyaev–Zeldovich Effect

“…search for connections between stars in the centers of galaxies and the morphology of those galaxies (see Tacchella et al (2019)), determining galactic parameters such as density, metallicity, surface brightness and ionization degree from galactic spectra emission lines in the visual, ultraviolet, infrared and submillimeter ranges (see Ucci et al (2017)). The methods of machine learning were also used to select 400 galaxies with high jet activity indices in the Sloan Digital Sky Survey spectra of two million galaxies (Baron and Poznanski, 2017).…”

Surveying the reach and maturity of machine learning and artificial intelligence in astronomy

“…Cabrera‐Vives et al () uncovered human biases that existed in morphological classification, which could be reduced through supervised ML. Other applications included predicting the HI content of galaxies based on optical observations (Rafieferantsoa, Andrianomena, & Davé, ), determining physical properties of galaxies from their emission‐line spectra (Ucci, Ferrara, Gallerani, & Pallottini, ), point source detection from radio interferometry surveys (Vafaei Sadr et al, ), and cross‐identification of sources from the Radio Galaxy Zoo (Alger et al, ). Training a CNN on mock images of rare “blue nugget” galaxies from cosmological simulations, such objects were successfully found in an observational sample from the CANDELS survey (Huertas‐Company et al, ). Distance measures .…”

CPz: Classification-aided photometric-redshift estimation

“…supervised and unsupervised methods are able to identify correlations, groupings and even outliers in vast datasets that would otherwise be impossible to visualize by a human eye. Many works have explored classification in astronomy using machinelearning techniques aiming towards separating stars from galaxies (Odewahn et al 1993;Soumagnac et al 2015), QSO identification (Brescia et al 2015), estimating physical parameters A&A 619, A14 (2018) (Ucci et al 2017), finding peculiar objects (Meusinger et al 2012), etc. In addition to the class of the object, multiwavelength information are useful in providing an estimate of the redshift as proposed by Baum (1962). All modern extragalactic surveys make extensive use of photometric redshift estimation as it is a cost-efficient method to determine distances of galaxies (COMBO-17, CFHTLS, CDFS and ECDFS, Lockman Hole, AEGIS, COSMOS, XXL, to name a few).…”