Dimensionality Reduction of SDSS Spectra with Variational Autoencoders

Abstract

High resolution galaxy spectra contain much information about galactic physics, but the high dimensionality of these spectra makes it difficult to fully utilize the information they contain. We apply variational autoencoders (VAEs), a non-linear dimensionality reduction technique, to a sample of spectra from the Sloan Digital Sky Survey. In contrast to Principal Component Analysis (PCA), a widely used technique, VAEs can capture non-linear relationships between latent parameters and the data. We find that a VAE can reconstruct the SDSS spectra well with only six latent parameters, outperforming PCA with the same number of components. Different galaxy classes are naturally separated in this latent space, without class labels having been given to the VAE. The VAE latent space is interpretable because the VAE can be used to make synthetic spectra at any point in latent space. For example, making synthetic spectra along tracks in latent space yields sequences of realistic spectra that interpolate between two different types of galaxies. Using the latent space to find outliers may yield interesting spectra: in our small sample, we immediately find unusual data artifacts and stars misclassified as galaxies. In this exploratory work, we show that VAEs create compact, interpretable latent spaces that capture non-linear features of the data. While a VAE takes substantial time to train (~1 day for 48000 spectra), once trained, VAEs can enable the fast exploration of large astronomical data sets.

arXiv comment22 pages, 17 figures, accepted to AJ; code available at https://github.com/stephenportillo/SDSS-VAE

References in corpus (32)

• Astropy: A Community Python Package for Astronomy
• The Sloan Digital Sky Survey: Technical Summary
• The Seventh Data Release of the Sloan Digital Sky Survey
• Binary companions of evolved stars in APOGEE DR14: Search method and catalog of ~5,000 companions
• Theoretical Modeling of Starburst Galaxies
• The 2.5 m Telescope of the Sloan Digital Sky Survey
• The Host Galaxies and Classification of Active Galactic Nuclei
• Spectroscopic Target Selection in the Sloan Digital Sky Survey: The Main Galaxy Sample
• The Multi-Object, Fiber-Fed Spectrographs for SDSS and the Baryon Oscillation Spectroscopic Survey
• Understanding Galaxy Evolution through Emission Lines
• Introduction to astroML: Machine Learning for Astrophysics
• Spectral Classification of Quasars in the Sloan Digital Sky Survey: Eigenspectra; Redshift and Luminosity Effects
• Distributions of Galaxy Spectral Types in the Sloan Digital Sky Survey
• Galaxy Types in the Sloan Digital Sky Survey Using Supervised Artificial Neural Networks
• The weirdest SDSS galaxies: results from an outlier detection algorithm
• A recurrent neural network for classification of unevenly sampled variable stars
• An Artificial Neural Network Approach to Classification of Galaxy Spectra
• Automatic unsupervised classification of all SDSS/DR7 galaxy spectra
• Ensemble Learning Independent Component Analysis of Normal Galaxy Spectra
• Painting with baryons: augmenting N-body simulations with gas using deep generative models
• A Machine Learning Based Morphological Classification of 14,245 Radio AGNs Selected From The Best-Heckman Sample
• Reducing the Dimensionality of Data: Locally Linear Embedding of Sloan Galaxy Spectra
• Unusual quasars from the Sloan Digital Sky Survey selected by means of Kohonen self-organising maps
• Detecting outliers and learning complex structures with large spectroscopic surveys - a case study with APOGEE stars
• Exploiting Low-Dimensional Structure in Astronomical Spectra
• A deep learning model to emulate simulations of cosmic reionization
• A large sample of Kohonen selected E+A (post-starburst) galaxies from the Sloan Digital Sky Survey
• An autoencoder of stellar spectra and its application in automatically estimating atmospheric parameters
• Deep Neural Network Classifier for Variable Stars with Novelty Detection Capability
• Parameterizing Stellar Spectra Using Deep Neural Networks
• X-ray Study of Spatial Structures in Tycho's Supernova Remnant Using Unsupervised Deep Learning
• Mapping the Similarities of Spectra: Global and Locally-biased Approaches to SDSS Galaxy Data

Cited by in corpus (40)

• The DAWES review 10: The impact of deep learning for the analysis of galaxy surveys
• AstroCLIP: A Cross-Modal Foundation Model for Galaxies
• Searching for changing-state AGNs in massive datasets -- I: applying deep learning and anomaly detection techniques to find AGNs with anomalous variability behaviours
• Accelerated Bayesian SED Modeling using Amortized Neural Posterior Estimation
• A variational encoder-decoder approach to precise spectroscopic age estimation for large Galactic surveys
• Data mining techniques on astronomical spectra data. II : Classification Analysis
• Applications of Machine Learning to Predicting Core-collapse Supernova Explosion Outcomes
• Autoencoding Galaxy Spectra I: Architecture
• Autoencoding Galaxy Spectra II: Redshift Invariance and Outlier Detection
• Stochastic Modelling of Star Formation Histories III. Constraints from Physically-Motivated Gaussian Processes
• Mapping the Diversity of Galaxy Spectra with Deep Unsupervised Machine Learning
• Interpreting automatic AGN classifiers with saliency maps
• Deriving the star formation histories of galaxies from spectra with simulation-based inference
• On the fundamentality of the radial acceleration relation for late-type galaxy dynamics
• De-noising of galaxy optical spectra with autoencoders
• A representation learning approach to probe for dynamical dark energy in matter power spectra
• Machine learning the gap between real and simulated nebulae: A domain-adaptation approach to classify ionised nebulae in nearby galaxies
• The weird and the wonderful in our Solar System: Searching for serendipity in the Legacy Survey of Space and Time
• KilonovAE: Exploring Kilonova Spectral Features with Autoencoders
• Investigation of stellar magnetic activity using variational autoencoder based on low-resolution spectroscopic survey
• Inferring redshift and galaxy properties via a multi-task neural net with probabilistic outputs: An application to simulated MOONS spectra
• Galaxy Spectra Networks (GaSNet). III. Generative pre-trained network for spectrum reconstruction, redshift estimate and anomaly detection
• Emission Line Predictions for Mock Galaxy Catalogues: a New Differentiable and Empirical Mapping from DESI
• On the application of dimensionality reduction and clustering algorithms for the classification of kinematic morphologies of galaxies
• A Deep Neural Network Based Reverse Radio Spectrogram Search Algorithm
• Bayesian decomposition of the Galactic multi-frequency sky using probabilistic autoencoders
• Data-Space Validation of High-Dimensional Models by Comparing Sample Quantiles
• An Unsupervised Machine Learning Approach to Identify Spectral Energy Distribution Outliers: Application to the S-PLUS DR4 data
• Stellar Karaoke: deep blind separation of terrestrial atmospheric effects out of stellar spectra by velocity whitening
• Galaxy Manifold: Characterizing and understanding galaxies with two parameters
• EmulART: Emulating Radiative Transfer -- A pilot study on autoencoder based dimensionality reduction for radiative transfer models
• Real-time Light Curve Classification Framework for the Wide Field Survey Telescope Using Modified Semi-supervised Variational Auto-Encoder
• Automated quasar continuum estimation using neural networks: a comparative study of deep-learning architectures
• Euclid preparation. LXVII. Deep learning true galaxy morphologies for weak lensing shear bias calibration
• Exploring Galaxy Properties of eCALIFA with Contrastive Learning
• A machine learning approach to assessing the presence of substructure in quasar host galaxies using the Hyper Suprime-Cam Subaru Strategic Program
• Interpreting anomaly detection of SDSS spectra
• Identifying Anomalous DESI Galaxy Spectra with a Variational Autoencoder
• Compression Method for Solar Polarization Spectra Collected from Hinode SOT/SP Observations
• J-PLUS: Identifying Outlier Spectral Energy Distributions in the Fourth Data Release Using Unsupervised Machine Learning