Machines Learn to Infer Stellar Parameters Just by Looking at a Large Number of Spectra

Abstract

Machine learning has been widely applied to clearly defined problems of astronomy and astrophysics. However, deep learning and its conceptual differences to classical machine learning have been largely overlooked in these fields. The broad hypothesis behind our work is that letting the abundant real astrophysical data speak for itself, with minimal supervision and no labels, can reveal interesting patterns which may facilitate discovery of novel physical relationships. Here as the first step, we seek to interpret the representations a deep convolutional neural network chooses to learn, and find correlations in them with current physical understanding. We train an encoder-decoder architecture on the self-supervised auxiliary task of reconstruction to allow it to learn general representations without bias towards any specific task. By exerting weak disentanglement at the information bottleneck of the network, we implicitly enforce interpretability in the learned features. We develop two independent statistical and information-theoretical methods for finding the number of learned informative features, as well as measuring their true correlation with astrophysical validation labels. As a case study, we apply this method to a dataset of ~270000 stellar spectra, each of which comprising ~300000 dimensions. We find that the network clearly assigns specific nodes to estimate (notions of) parameters such as radial velocity and effective temperature without being asked to do so, all in a completely physics-agnostic process. This supports the first part of our hypothesis. Moreover, we find with high confidence that there are ~4 more independently informative dimensions that do not show a direct correlation with our validation parameters, presenting potential room for future studies.

References in corpus (25)

• The information bottleneck method
• The SIMBAD astronomical database
• The VizieR database of Astronomical Catalogues
• Tutorial on Variational Autoencoders
• The Revised TESS Input Catalog and Candidate Target List
• Physics Informed Deep Learning (Part I): Data-driven Solutions of Nonlinear Partial Differential Equations
• Equitability, mutual information, and the maximal information coefficient
• PPINN: Parareal Physics-Informed Neural Network for time-dependent PDEs
• Discovering physical concepts with neural networks
• Recent Advances in Autoencoder-Based Representation Learning
• InfoVAE: Information Maximizing Variational Autoencoders
• The HARPS survey for southern extra-solar planets II. A 14 Earth-masses exoplanet around mu Arae
• Learning New Physics from a Machine
• The weirdest SDSS galaxies: results from an outlier detection algorithm
• Guiding New Physics Searches with Unsupervised Learning
• Photometric redshifts with the Multilayer Perceptron Neural Network: application to the HDF-S and SDSS
• Automated Classification of Stellar Spectra. II: Two-Dimensional Classification with Neural Networks and Principal Components Analysis
• Physics enhanced neural networks predict order and chaos
• Photometry of high-redshift blended galaxies using deep learning
• Learning to Perform Physics Experiments via Deep Reinforcement Learning
• Effective Image Differencing with ConvNets for Real-time Transient Hunting
• Machine-assisted discovery of relationships in astronomy
• Learning A Physical Long-term Predictor
• The first analytical expression to estimate photometric redshifts suggested by a machine
• Hybrid Learning of Optical Flow and Next Frame Prediction to Boost Optical Flow in the Wild

Cited by in corpus (6)

• A robust estimator of mutual information for deep learning interpretability
• Explaining dark matter halo density profiles with neural networks
• A representation learning approach to probe for dynamical dark energy in matter power spectra
• Investigation of stellar magnetic activity using variational autoencoder based on low-resolution spectroscopic survey
• Stellar Karaoke: deep blind separation of terrestrial atmospheric effects out of stellar spectra by velocity whitening
• Discovering the building blocks of dark matter halo density profiles with neural networks