Using convolutional neural networks to predict galaxy metallicity from three-color images

Abstract

We train a deep residual convolutional neural network (CNN) to predict the gas-phase metallicity ($Z$) of galaxies derived from spectroscopic information ($Z\equiv 12+\log (O/H)$) using only three-band $gri$ images from the Sloan Digital Sky Survey. When trained and tested on $128\times 128$-pixel images, the root mean squared error (RMSE) of $Z_{\mathrm{pred}}-Z_{\mathrm{true}}$ is only 0.085 dex, vastly outperforming a trained random forest algorithm on the same data set (RMSE $=0.130$ dex). The amount of scatter in $Z_{\mathrm{pred}}-Z_{\mathrm{true}}$ decreases with increasing image resolution in an intuitive manner. We are able to use CNN-predicted $Z_{\mathrm{pred}}$ and independently measured stellar masses to recover a mass-metallicity relation with $0.10$ dex scatter. Because our predicted MZR shows no more scatter than the empirical MZR, the difference between $Z_{\mathrm{pred}}$ and $Z_{\mathrm{true}}$ can not be due to purely random error. This suggests that the CNN has learned a representation of the gas-phase metallicity, from the optical imaging, beyond what is accessible with oxygen spectral lines.

arXiv comment13 pages, 6 figures, accepted to MNRAS. Code is available at https://github.com/jwuphysics/galaxy-cnns

References in corpus (50)

• Adam: A Method for Stochastic Optimization
• Very Deep Convolutional Networks for Large-Scale Image Recognition
• Scikit-learn: Machine Learning in Python
• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• The Sloan Digital Sky Survey: Technical Summary
• Decoupled Weight Decay Regularization
• On the variation of the Initial Mass Function
• Improving neural networks by preventing co-adaptation of feature detectors
• Deep Residual Learning for Image Recognition
• The Origin of the Mass--Metallicity Relation: Insights from 53,000 Star-Forming Galaxies in the SDSS
• The physical properties of star forming galaxies in the low redshift universe
• Stellar Masses and Star Formation Histories for 10^5 Galaxies from the Sloan Digital Sky Survey
• SGDR: Stochastic Gradient Descent with Warm Restarts
• UV Star Formation Rates in the Local Universe
• COSMOS Photometric Redshifts with 30-bands for 2-deg2
• A fundamental relation between mass, SFR and metallicity in local and high redshift galaxies
• Gas-regulation of galaxies: the evolution of the cosmic sSFR, the metallicity-mass-SFR relation and the stellar content of haloes
• Rotation-invariant convolutional neural networks for galaxy morphology prediction
• The Fourteenth Data Release of the Sloan Digital Sky Survey: First Spectroscopic Data from the extended Baryon Oscillation Spectroscopic Survey and from the second phase of the Apache Point Observatory Galactic Evolution Experiment
• An Analytic Model for the Evolution of the Stellar, Gas, and Metal Content of Galaxies
• Preparing Red-Green-Blue (RGB) Images from CCD Data
• Star formation rate indicators in the Sloan Digital Sky Survey
• A Critical Assessment of Photometric Redshift Methods: A CANDELS Investigation
• The Metallicity-Luminosity Relation, Effective Yields, and Metal Loss in Spiral and Irregular Galaxies
• A catalog of visual-like morphologies in the 5 CANDELS fields using deep-learning
• The Mass-Metallicity relation explored with CALIFA: I. Is there a dependence on the star formation rate?
• Fast Automated Analysis of Strong Gravitational Lenses with Convolutional Neural Networks
• The Origin and Evolution of the Mass-Metallicity Relationship for Galaxies: Results from Cosmological N-Body Simulations
• Star-galaxy Classification Using Deep Convolutional Neural Networks
• CMU DeepLens: Deep Learning For Automatic Image-based Galaxy-Galaxy Strong Lens Finding
• Photometric redshifts from SDSS images using a Convolutional Neural Network
• Finding Strong Gravitational Lenses in the Kilo Degree Survey with Convolutional Neural Networks
• The Formation of Dust Lanes: Implications for Galaxy Evolution
• Fuel Efficient Galaxies: Sustaining Star Formation with Stellar Mass Loss
• The Chemical Evolution of Star-Forming Galaxies Over the Last 11 Billion Years
• Photometric redshift estimation via deep learning
• Using transfer learning to detect galaxy mergers
• An Application of Deep Neural Networks in the Analysis of Stellar Spectra
• Galaxy Zoo: Quantitative Visual Morphological Classifications for 48,000 galaxies from CANDELS
• An automatic taxonomy of galaxy morphology using unsupervised machine learning
• Deep Learning Identifies High-z Galaxies in a Central Blue Nugget Phase in a Characteristic Mass Range
• A Machine Learning Approach for Dynamical Mass Measurements of Galaxy Clusters
• On the Mass-Metallicity-Star Formation Rate Relation for Galaxies at $z\sim 2$
• Integrating human and machine intelligence in galaxy morphology classification tasks
• CLASH: Accurate Photometric Redshifts with 14 HST bands in Massive Galaxy Cluster Cores
• Testing Convolutional Neural Networks for finding strong gravitational lenses in KiDS
• A First Look at creating mock catalogs with machine learning techniques
• How to measure metallicity from five-band photometry with supervised machine learning algorithms
• The Velocity Distribution Function of Galaxy Clusters as a Cosmological Probe
• Visualizing the Hidden Features of Galaxy Morphology with Machine Learning

Cited by in corpus (18)

• Automatic detection of low surface brightness galaxies from SDSS images
• Prospects for Recovering Galaxy Intrinsic Shapes from Projected Quantities
• Comparison of Observed Galaxy Properties with Semianalytic Model Predictions using Machine Learning
• Identification of Galaxy-Galaxy Strong Lens Candidates in the DECam Local Volume Exploration Survey Using Machine Learning
• Connecting optical morphology, environment, and HI mass fraction for low-redshift galaxies using deep learning
• Tracing Halpha Fibrils through Bayesian Deep Learning
• Identifying and Tracking Solar Magnetic Flux Elements with Deep Learning
• Euclid Preparation XXXIII. Characterization of convolutional neural networks for the identification of galaxy-galaxy strong lensing events
• Constraining stellar population parameters from narrow band photometric surveys using convolutional neural networks
• Predicting galaxy spectra from images with hybrid convolutional neural networks
• Inferring redshift and galaxy properties via a multi-task neural net with probabilistic outputs: An application to simulated MOONS spectra
• Interpreting deep learning-based stellar mass estimation via causal analysis and mutual information decomposition
• Insights on Galaxy Evolution from Interpretable Sparse Feature Networks
• SN 2017fgc: A Fast-Expanding Type Ia Supernova Exploded in Massive Shell Galaxy NGC 474
• Galaxy And Mass Assembly (GAMA): From filaments to voids, how extreme environment affects gas metallicity and SFR in galaxies
• Photometric Redshift Estimation with Convolutional Neural Networks and Galaxy Images: A Case Study of Resolving Biases in Data-Driven Methods
• A Value-added Physical Properties Catalog for Low-redshift Galaxies from DESI Legacy Imaging Surveys DR10
• Extending the SAGA Survey (xSAGA) I: Satellite Radial Profiles as a Function of Host Galaxy Properties