Predicting resolved galaxy properties from photometric images using convolutional neural networks

Abstract

Multi-band images of galaxies reveal a huge amount of information about their morphology and structure. However, inferring properties of the underlying stellar populations such as age, metallicity or kinematics from those images is notoriously difficult. Traditionally such information is best extracted from expensive spectroscopic observations. Here we present the Painting IntrinsiC Attributes onto SDSS Objects (PICASSSO) project and test the information content of photometric multi-band images of galaxies. We train a convolutional neural network on 27,558 galaxy image pairs to establish a connection between broad-band images and the underlying physical stellar and gaseous galaxy property maps. We test our machine learning (ML) algorithm with SDSS ugriz mock images for which uncertainties and systematics are exactly known. We show that multi-band galaxy images contain enough information to reconstruct 2d maps of stellar mass, metallicity, age and gas mass, metallicity as well as star formation rate. We recover the true stellar properties on a pixel by pixel basis with only little scatter, ∼ < 20% compared to ∼ 50% statistical uncertainty from traditional mass-to-light-ratio based methods. We further test for any systematics of our algorithm with image resolution, training sample size or wavelength coverage. We find that galaxy morphology alone constrains stellar properties to better than ∼ 20% thus highlighting the benefits of including morphology into the parameter estimation. The machine learning approach can predict maps of high resolution, only limited by the resolution of the input bands, thus achieve higher resolution than IFU observations. The network architecture and all code is publicly available on GitHub .

Harnessing the Hubble Space Telescope Archives: A Catalog of 21,926 Interacting Galaxies

“…The workhorse algorithm is the convolutional neural network (CNN; for an introduction, see O 'Shea & Nash 2015), most often used in image recognition and feature extraction. CNNs can be used for general classification (e.g., early-versus late-type galaxies) or to extract specific morphological features of galaxies, such as bars and spiral arms; many works have demonstrated their effectiveness at this (e.g., Ackermann et al 2018;Jacobs et al 2019;Bickley et al 2021;Buck & Wolf 2021;Walmsley et al 2022a). Pearson et al (2022) demonstrated the power of CNNs for finding interacting and merging galaxies specifically, finding 2109 in 5.4 deg 2 of Hyper Suprime-Cam imagery-a large sample for the small area covered.…”

Galaxy Light Profile Convolutional Neural Networks (GaLNets). I. Fast and Accurate Structural Parameters for Billion-galaxy Samples

“…CNNs have also been successfully applied to a variety of galaxy photometry-related studies, such as nonparametric light profile extraction (e.g., Smith et al 2021;Stone et al 2021), source deblending (e.g., Boucaud et al 2020), and galaxy stellar population analysis (e.g., Buck & Wolf 2021). Tuccillo et al (2018) for the first time applied CNNs to twodimensional light profile galaxy fitting on Hubble Space Telescope (HST)/CANDELS data.…”

Using Artificial Intelligence and real galaxy images to constrain parameters in galaxy formation simulations

“…Such ML techniques have been applied in a large number of astronomical use-cases related to galaxy properties (e.g. Dieleman et al 2015;Beck et al 2018;Hocking et al 2018;Dawson et al 2020;Buck & Wolf 2021).…”