Deep learning of multi-element abundances from high-resolution spectroscopic data

Abstract

Deep learning with artificial neural networks is increasingly gaining attention, because of its potential for data-driven astronomy. However, this methodology usually does not provide uncertainties and does not deal with incompleteness and noise in the training data. In this work, we design a neural network for high-resolution spectroscopic analysis using APOGEE data that mimics the methodology of standard spectroscopic analyses: stellar parameters are determined using the full wavelength range, but individual element abundances use censored portions of the spectrum. We train this network with a customized objective function that deals with incomplete and noisy training data and apply dropout variational inference to derive uncertainties on our predictions. We determine parameters and abundances for 18 individual elements at the $\approx 0.03$ dex level, even at low signal-to-noise ratio. We demonstrate that the uncertainties returned by our method are a realistic estimate of the precision and they automatically blow up when inputs or outputs outside of the training set are encountered, thus shielding users from unwanted extrapolation. By using standard deep-learning tools for GPU acceleration, our method is extremely fast, allowing analysis of the entire APOGEE data set of $\approx 250,000$ spectra in ten minutes on a single, low-cost GPU. We release the stellar parameters and 18 individual-element abundances with associated uncertainty for the entire APOGEE DR14 dataset. Simultaneously, we release astroNN, a well-tested, open-source python package developed for this work, but that is also designed to be a general package for deep learning in astronomy. astroNN is available at https://github.com/henrysky/astroNN with extensive documentation at http://astroNN.readthedocs.io.

arXiv commentAccepted 08 November 2018 to MNRAS. astroNN available at https://github.com/henrysky/astroNN and paper-specific code available at https://github.com/henrysky/astroNN_spectra_paper_figures

References in corpus (20)

• TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems
• The Astropy Project: Building an inclusive, open-science project and status of the v2.0 core package
• Improving neural networks by preventing co-adaptation of feature detectors
• PARSEC: stellar tracks and isochrones with the PAdova and TRieste Stellar Evolution Code
• What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?
• Galaxy Zoo : Morphologies derived from visual inspection of galaxies from the Sloan Digital Sky Survey
• cuDNN: Efficient Primitives for Deep Learning
• Euclid Definition Study Report
• LSST Science Book, Version 2.0
• A Berkeley View of Systems Challenges for AI
• An Application of Deep Neural Networks in the Analysis of Stellar Spectra
• SDSS-V: Pioneering Panoptic Spectroscopy
• Physical Parameterization of Stellar Spectra: The Neural Network Approach
• The Cannon 2: A data-driven model of stellar spectra for detailed chemical abundance analyses
• Detecting outliers and learning complex structures with large spectroscopic surveys - a case study with APOGEE stars
• The dimensionality of stellar chemical space using spectra from the Apache Point Observatory Galactic Evolution Experiment
• The Detailed Science Case for the Maunakea Spectroscopic Explorer: the Composition and Dynamics of the Faint Universe
• An autoencoder of stellar spectra and its application in automatically estimating atmospheric parameters
• LAMOST Spectral Survey
• The Peculiar Globular Cluster Palomar 1 and Persistence in the SDSS-APOGEE Database

Cited by in corpus (107)

• The Sixteenth Data Release of the Sloan Digital Sky Surveys: First Release from the APOGEE-2 Southern Survey and Full Release of eBOSS Spectra
• Painting a portrait of the Galactic disc with its stellar clusters
• Evidence for Two Early Accretion Events That Built the Milky Way Stellar Halo
• The Payne: self-consistent ab initio fitting of stellar spectra
• Evidence from APOGEE for the presence of a major building block of the halo buried in the inner Galaxy
• Dynamical heating across the Milky Way disc using APOGEE and $Gaia$
• Life in the fast lane: a direct view of the dynamics, formation, and evolution of the Milky Way's bar
• Abundance Estimates for 16 Elements in 6 Million Stars from LAMOST DR5 Low-Resolution Spectra
• Confirmation of the $Gaia$ DR2 parallax zero-point offset using asteroseismology and spectroscopy in the $Kepler$ field
• Simultaneous calibration of spectro-photometric distances and the Gaia DR2 parallax zero-point offset with deep learning
• Chemodynamics of barred galaxies in cosmological simulations: On the Milky Way's quiescent merger history and in-situ bulge
• APOGEE Chemical Abundance Patterns of the Massive Milky Way Satellites
• The chemical characterisation of halo substructure in the Milky Way based on APOGEE
• Weighing the stellar constituents of the Galactic halo with APOGEE red giant stars
• 3D NLTE spectral line formation of lithium in late-type stars
• How Many Components? Quantifying the Complexity of the Metallicity Distribution in the Milky Way Bulge with APOGEE
• Stellar migrations and metal flows -- Chemical evolution of the thin disc of a simulated Milky Way analogous galaxy
• A Tale of Two Disks: Mapping the Milky Way with the Final Data Release of APOGEE
• A direct measurement of the distance to the Galactic center using the kinematics of bar stars
• SPCANet: Stellar Parameters and Chemical Abundances Network for LAMOST-II Medium Resolution Survey
• The age-chemical abundance structure of the Galactic disc II: $\alpha$-dichotomy and thick disc formation
• Unveiling the time evolution of chemical abundances across the Milky Way disk with APOGEE
• The Pristine Inner Galaxy Survey (PIGS) II: Uncovering the most metal-poor populations in the inner Milky Way
• The Gaia DR2 parallax zero point: Hierarchical modeling of red clump stars
• The Relationship Between Globular Cluster Mass, Metallicity, and Light Element Abundance Variations
• BACCHUS Analysis of Weak Lines in APOGEE Spectra (BAWLAS)
• How Many Elements Matter?
• Stellar Abundance Maps of the Milky Way Disk
• Exploring the stellar age distribution of the Milky Way Bulge using APOGEE
• The contribution of N-rich stars to the Galactic stellar halo using APOGEE red giants
• Chemical Cartography with APOGEE: Mapping Disk Populations with a Two-Process Model and Residual Abundances
• The Kinematic Properties of Milky Way Stellar Halo Populations
• APOGEE Net: Improving the derived spectral parameters for young stars through deep learning
• Lithium Enrichment Signatures of Planetary Engulfment Events in Evolved Stars
• The Milky Way tomography with APOGEE: intrinsic density distribution and structure of mono-abundance populations
• Quantifying radial migration in the Milky Way: Inefficient over short timescales but essential to the very outer disc beyond ~15 kpc
• The Spectroscopic Binaries from LAMOST Medium-Resolution Survey (MRS). I. Searching for Double-lined Spectroscopic Binaries (SB2s) with Convolutional Neural Network
• Towards an astronomical foundation model for stars with a Transformer-based model
• Abundances in the Milky Way across five nucleosynthetic channels from 4 million LAMOST stars
• The CARMENES search for exoplanets around M dwarfs -- A deep learning approach to determine fundamental parameters of target stars
• Strong chemical tagging with APOGEE: 21 candidate star clusters that have dissolved across the Milky Way disc
• The GALAH Survey: Chemical Clocks
• A variational encoder-decoder approach to precise spectroscopic age estimation for large Galactic surveys
• CAPOS: The bulge Cluster APOgee Survey I. Overview and initial ASPCAP results
• The distribution of [$\alpha$/Fe] in the Milky Way disc
• Blind chemical tagging with DBSCAN: prospects for spectroscopic surveys
• Cycle-StarNet: Bridging the gap between theory and data by leveraging large datasets
• The RAdial Velocity Experiment (RAVE): Parameterisation of RAVE spectra based on convolutional neural networks
• Combined APOGEE-GALAH stellar catalogues using the Cannon
• The stellar parameters and elemental abundances from low-resolution spectra I: 1.2 million giants from LAMOST DR8
• Small Planets in the Galactic Context: Host Star Kinematics, Iron, and $\alpha$ Element Enhancement
• Assessing the performance of LTE and NLTE synthetic stellar spectra in a machine learning framework
• Computational Tools for the Spectroscopic Analysis of White Dwarfs
• Estimating Stellar Parameters from LAMOST Low-resolution Spectra
• The APO-K2 Catalog. II. Accurate Stellar Ages for Red Giant Branch Stars across the Milky Way
• Neural network-based anomaly detection for high-resolution X-ray spectroscopy
• Universal properties of the high- and low-α disk: small intrinsic abundance scatter and migrating stars
• Disentangled Representation Learning for Astronomical Chemical Tagging
• Identify Light-Curve Signals with Deep Learning Based Object Detection Algorithm. I. Transit Detection
• An enquiry on the origins of N-rich stars in the inner Galaxy basedon APOGEE chemical compositions
• deepSIP: Linking Type Ia Supernova Spectra to Photometric Quantities with Deep Learning
• Tracing Halpha Fibrils through Bayesian Deep Learning
• A data-driven model of nucleosynthesis with chemical tagging in a lower-dimensional latent space
• DIGS: Deep Inference of Galaxy Spectra with Neural Posterior Estimation
• Identifying and Tracking Solar Magnetic Flux Elements with Deep Learning
• Searching for Solar Siblings in APOGEE and $Gaia$ DR2 with N-body Simulations
• Searching for the extra-tidal stars of globular clusters using high-dimensional analysis and a core particle spray code
• Testing the chemical homogeneity of chemically tagged dissolved birth clusters
• Estimating Atmospheric Parameters from LAMOST Low-Resolution Spectra with Low SNR
• The APOGEE Library of Infrared SSP Templates (A-LIST): High-Resolution Simple Stellar Population Spectral Models in the H-Band
• Elemental abundances distributions in (R, $V_{\varphi }$) plane with LAMOST DR5 and Gaia DR2
• Chemical Cartography with APOGEE: Two-process Parameters and Residual Abundances for 288,789 Stars from Data Release 17
• The Swan: Data-Driven Inference of Stellar Surface Gravities for Cool Stars from Photometric Light Curves
• A Model RRNet for Spectral Information Exploitation and LAMOST Medium-resolution Spectrum Parameter Estimation
• The chemical enrichment of the Milky Way disk evaluated using conditional abundances
• An unsupervised method for identifying $X$-enriched stars directly from spectra: Li in LAMOST
• Is Terzan 5 the remnant of a building block of the Galactic bulge? Evidence from APOGEE
• Dark energy reconstruction analysis with artificial neural networks: Application on simulated Supernova Ia data from Rubin Observatory
• Using Bayesian Deep Learning to infer Planet Mass from Gaps in Protoplanetary Disks
• Galaxy Spectra neural Network (GaSNet). II. Using Deep Learning for Spectral Classification and Redshift Predictions
• Li-rich Giants Identified from LAMOST DR8 Low-Resolution Survey
• A machine learning approach to photometric metallicities of giant stars
• Reliable stellar abundances of individual stars with the MUSE integral-field spectrograph
• Physics-Integrated Variational Autoencoders for Robust and Interpretable Generative Modeling
• Closing the stellar labels gap: Stellar label independent evidence for [$\alpha /M$] information in Gaia BP/RP spectra
• L dwarfs detection from SDSS images using improved Faster R-CNN
• Are Milky-Way-like galaxies like the Milky Way? A view from SDSS-IV/MaNGA
• Galaxy Spectra Networks (GaSNet). III. Generative pre-trained network for spectrum reconstruction, redshift estimate and anomaly detection
• SSSpaNG! Stellar Spectra as Sparse, data-driven, Non-Gaussian processes
• A Galactic Self-Portrait: Density Structure and Integrated Properties of the Milky Way Disk
• Classifying Seyfert galaxies with deep learning
• Mapping the Milky Way in 5-D with 170 Million Stars
• The Seventeenth Data Release of the Sloan Digital Sky Surveys: Complete Release of MaNGA, MaStar and APOGEE-2 Data
• Magnesium isotope ratios in Milky Way and dwarf galaxy stars
• Weighing stars from birth to death: mass determination methods across the HRD
• StarUnLink: identifying and mitigating signals from communications satellites in stellar spectral surveys
• A Bayesian neural network predicts the dissolution of compact planetary systems
• Disentanglement of the chemodynamical assembly: mapping the Milky Way disks
• Astronomical Images Quality Assessment with Automated Machine Learning
• On the connection between nitrogen-enhanced field stars and the Galactic globular clusters
• PSF-based Analysis for Detecting Unresolved Wide Binaries
• 200,000+ Deep Learning-inferred Periods of Stellar Variability from the All-Sky Automated Survey for Supernovae
• Neutron-capture elements record the ordered chemical evolution of the disc over time
• Functional Data Analysis for Extracting the Intrinsic Dimensionality of Spectra: Application to Chemical Homogeneity in the Open Cluster M67
• Inactive longitude and superflare in the active single-lined pre-main sequence binary V2279 Cyg
• Exploiting Redundancy: Separable Group Convolutional Networks on Lie Groups
• SF-R You Sure? The Conflicting Role of Star Formation Rates in Constraining the Evolution of Milky Way Analogues in Cosmological Simulations