Abstract:We present starduster, a supervised deep-learning model that predicts the multiwavelength spectral energy distribution (SED) from galaxy geometry parameters and star formation history by emulating dust radiative transfer simulations. The model is composed of three specifically designed neural networks, which take into account the features of dust attenuation and emission. We utilize the skirt radiative transfer simulation to produce data for the training data of neural networks. Each neural network can be trai… Show more
“…At the time of writing, there are many state-of-the-art SEDfitting tools. To give some examples among the most recently developed, we cite AGNfitter (Calistro Rivera et al 2016), BEAGLE (Chevallard & Charlot 2016), Pipe3D (Sánchez et al 2016), Prospector (Leja et al 2017;Johnson et al 2021), Dense Basis (Iyer & Gawiser 2017;Iyer et al 2019), FIREFLY (Wilkinson et al 2017), Lightning (Eufrasio 2017), Mr-Moose (Drouart & Falkendal 2018), BAGPIPES (Carnall et al 2018), FortesFit (Rosario 2019), PEGASE.3 (Fioc & Rocca-Volmerange 2019), X-CIGALE (Yang et al 2020), MCSED (Bowman et al 2020), MIRKWOOD (Gilda et al 2021), piXedfit (Abdurro'uf et al 2021), ProSpect (Thorne et al 2021), and Starduster (Qiu & Kang 2022). Machine-learning techniques are also advancing rapidly to provide redshift and physical parameter estimates for galaxies across a large redshift range, e.g., Davidzon et al (2019) and Simet et al (2021).…”
The study of galaxy evolution hinges on our ability to interpret multiwavelength galaxy observations in terms of their physical properties. To do this, we rely on spectral energy distribution (SED) models, which allow us to infer physical parameters from spectrophotometric data. In recent years, thanks to wide and deep multiwave band galaxy surveys, the volume of high-quality data have significantly increased. Alongside the increased data, algorithms performing SED fitting have improved, including better modeling prescriptions, newer templates, and more extensive sampling in wavelength space. We present a comprehensive analysis of different SED-fitting codes including their methods and output with the aim of measuring the uncertainties caused by the modeling assumptions. We apply 14 of the most commonly used SED-fitting codes on samples from the CANDELS photometric catalogs at z ∼ 1 and z ∼ 3. We find agreement on the stellar mass, while we observe some discrepancies in the star formation rate (SFR) and dust-attenuation results. To explore the differences and biases among the codes, we explore the impact of the various modeling assumptions as they are set in the codes (e.g., star formation histories, nebular, dust and active galactic nucleus models) on the derived stellar masses, SFRs, and A
V
values. We then assess the difference among the codes on the SFR–stellar mass relation and we measure the contribution to the uncertainties by the modeling choices (i.e., the modeling uncertainties) in stellar mass (∼0.1 dex), SFR (∼0.3 dex), and dust attenuation (∼0.3 mag). Finally, we present some resources summarizing best practices in SED fitting.
“…At the time of writing, there are many state-of-the-art SEDfitting tools. To give some examples among the most recently developed, we cite AGNfitter (Calistro Rivera et al 2016), BEAGLE (Chevallard & Charlot 2016), Pipe3D (Sánchez et al 2016), Prospector (Leja et al 2017;Johnson et al 2021), Dense Basis (Iyer & Gawiser 2017;Iyer et al 2019), FIREFLY (Wilkinson et al 2017), Lightning (Eufrasio 2017), Mr-Moose (Drouart & Falkendal 2018), BAGPIPES (Carnall et al 2018), FortesFit (Rosario 2019), PEGASE.3 (Fioc & Rocca-Volmerange 2019), X-CIGALE (Yang et al 2020), MCSED (Bowman et al 2020), MIRKWOOD (Gilda et al 2021), piXedfit (Abdurro'uf et al 2021), ProSpect (Thorne et al 2021), and Starduster (Qiu & Kang 2022). Machine-learning techniques are also advancing rapidly to provide redshift and physical parameter estimates for galaxies across a large redshift range, e.g., Davidzon et al (2019) and Simet et al (2021).…”
The study of galaxy evolution hinges on our ability to interpret multiwavelength galaxy observations in terms of their physical properties. To do this, we rely on spectral energy distribution (SED) models, which allow us to infer physical parameters from spectrophotometric data. In recent years, thanks to wide and deep multiwave band galaxy surveys, the volume of high-quality data have significantly increased. Alongside the increased data, algorithms performing SED fitting have improved, including better modeling prescriptions, newer templates, and more extensive sampling in wavelength space. We present a comprehensive analysis of different SED-fitting codes including their methods and output with the aim of measuring the uncertainties caused by the modeling assumptions. We apply 14 of the most commonly used SED-fitting codes on samples from the CANDELS photometric catalogs at z ∼ 1 and z ∼ 3. We find agreement on the stellar mass, while we observe some discrepancies in the star formation rate (SFR) and dust-attenuation results. To explore the differences and biases among the codes, we explore the impact of the various modeling assumptions as they are set in the codes (e.g., star formation histories, nebular, dust and active galactic nucleus models) on the derived stellar masses, SFRs, and A
V
values. We then assess the difference among the codes on the SFR–stellar mass relation and we measure the contribution to the uncertainties by the modeling choices (i.e., the modeling uncertainties) in stellar mass (∼0.1 dex), SFR (∼0.3 dex), and dust attenuation (∼0.3 mag). Finally, we present some resources summarizing best practices in SED fitting.
“…Deschamps et al 2015;Hendrix et al 2016;Mosenkov et al 2021;Ebenbichler et al 2022;Jáquez-Domínguez et al 2023), SKIRT is mainly used in an extragalactic context. It is used to generate images, spectra, spectral energy distributions, and polarisation maps for idealised galaxy models (Gadotti et al 2010;De Geyter et al 2014;Lee et al 2016;Lin et al 2021;Qiu & Kang 2022), for high-resolution models fitted to observed spiral galaxies (De Looze et al 2014;Viaene et al 2017Viaene et al , 2020Mosenkov et al 2018;Williams et al 2019;Verstocken et al 2020;Nersesian et al 2020a,b), and for simulated galaxies extracted from cosmological simulations (e.g. Saftly et al 2015;Trayford et al 2017;Rodriguez-Gomez et al 2019;Vogelsberger et al 2020b;Liang et al 2021;Hsu et al 2023;Cochrane et al 2023).…”
Section: The Skirt Radiative Transfer Codementioning
Galaxy morphology is a powerful diagnostic to assess the realism of cosmological hydrodynamical simulations. Determining the morphology of simulated galaxies requires the generation of synthetic images through 3D radiative transfer post-processing that properly accounts for different stellar populations and interstellar dust attenuation. We use the SKIRT code to generate the TNG50-SKIRT Atlas, a synthetic UV to near-infrared broadband image atlas for a complete stellar-mass selected sample of 1154 galaxies extracted from the TNG50 cosmological simulation at $z=0$. The images have a high spatial resolution (100 pc) and a wide field of view (160 kpc). In addition to the dust-obscured images, we also release dust-free images and physical parameter property maps with matching characteristics. As a sanity check and preview application we discuss the UVJ diagram of the galaxy sample. We investigate the effect of dust attenuation on the UVJ diagram and find that it affects both the star-forming and the quiescent galaxy populations. The quiescent galaxy region is polluted by younger and star-forming highly inclined galaxies, while dust attenuation induces a separation in inclination of the star-forming galaxy population, with low-inclination galaxies remaining at the blue side of the diagram and high-inclination galaxies systematically moving towards the red side. This image atlas can be used for a variety of other applications, including galaxy morphology studies and the investigation of local scaling relations. We publicly release the images and parameter maps, and we invite the community to use them.
“…The authors of this first work show a reasonable reconstruction of the SFH and a decent robustness to domain changes. Qiu & Kang (2021) uses CNNs for the opposite task, that is, estimate the galaxy SED from the galaxy SFH taken from simulations. In this case deep learning acts as an emulator of radiative transfer codes. Figure 24. CNN applied to reconstruct the Star Formation Histories of galaxies.…”
Section: Deep Learning For Inferring Physical Properties Of Galaxiesmentioning
The amount and complexity of data delivered by modern galaxy surveys has been steadily increasing over the past years. New facilities will soon provide imaging and spectra of hundreds of millions of galaxies. Extracting coherent scientific information from these large and multi-modal data sets remains an open issue for the community and data-driven approaches such as deep learning have rapidly emerged as a potentially powerful solution to some long lasting challenges. This enthusiasm is reflected in an unprecedented exponential growth of publications using neural networks, which have gone from a handful of works in 2015 to an average of one paper per week in 2021 in the area of galaxy surveys. Half a decade after the first published work in astronomy mentioning deep learning, and shortly before new big data sets such as Euclid and LSST start becoming available, we believe it is timely to review what has been the real impact of this new technology in the field and its potential to solve key challenges raised by the size and complexity of the new datasets. The purpose of this review is thus two-fold. We first aim at summarising, in a common document, the main applications of deep learning for galaxy surveys that have emerged so far. We then extract the major achievements and lessons learned and highlight key open questions and limitations, which in our opinion, will require particular attention in the coming years. Overall, state-of-the-art deep learning methods are rapidly adopted by the astronomical community, reflecting a democratisation of these methods. This review shows that the majority of works using deep learning up to date are oriented to computer vision tasks (e.g. classification, segmentation). This is also the domain of application where deep learning has brought the most important breakthroughs so far. However, we also report that the applications are becoming more diverse and deep learning is used for estimating galaxy properties, identifying outliers or constraining the cosmological model. Most of these works remain at the exploratory level though which could partially explain the limited impact in terms of citations. Some common challenges will most likely need to be addressed before moving to the next phase of massive deployment of deep learning in the processing of future surveys; for example, uncertainty quantification, interpretability, data labelling and domain shift issues from training with simulations, which constitutes a common practice in astronomy.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
