Bayesian decomposition of the Galactic multi-frequency sky using probabilistic autoencoders

Context. All-sky observations show both Galactic and non-Galactic diffuse emission, for example from interstellar matter or the cosmic microwave background (CMB). The decomposition of the emission into different underlying radiative components is an important signal reconstruction problem. Aims. We aim to reconstruct radiative all-sky components using spectral data, without incorporating knowledge about physical or spatial correlations. Methods. We built a self-instructing algorithm based on variational autoencoders following three steps: (1)We stated a forward model describing how the data set was generated from a smaller set of features, (2) we used Bayes’ theorem to derive a posterior probability distribution, and (3) we used variational inference and statistical independence of the features to approximate the posterior. From this, we derived a loss function and optimized it with neural networks. The resulting algorithm contains a quadratic error norm with a self-adaptive variance estimate to minimize the number of hyperparameters. We trained our algorithm on independent pixel vectors, each vector representing the spectral information of the same pixel in 35 Galactic all-sky maps ranging from the radio to the γ-ray regime. Results. The algorithm calculates a compressed representation of the input data. We find the feature maps derived in the algorithm’s latent space show spatial structures that can be associated with all-sky representations of known astrophysical components. Our resulting feature maps encode (1) the dense interstellar medium (ISM), (2) the hot and dilute regions of the ISM, and (3) the CMB, without being informed about these components a priori. Conclusions. We conclude that Bayesian signal reconstruction with independent Gaussian latent space statistics is sufficient to reconstruct the dense and the dilute ISM, as well as the CMB, from spectral correlations only. The computational approximation of the posterior can be performed efficiently using variational inference and neural networks, making them a suitable approach to probabilistic data analysis.

[1]  R. B. Barreiro,et al.  Planck2018 results , 2020, Astronomy & Astrophysics.

[2]  A. Connolly,et al.  Dimensionality Reduction of SDSS Spectra with Variational Autoencoders , 2020, The Astronomical Journal.

[3]  J. Xavier Prochaska,et al.  Effectively using unsupervised machine learning in next generation astronomical surveys , 2019, Astron. Comput..

[4]  Giuseppe Longo,et al.  Foreword to the Focus Issue on Machine Intelligence in Astronomy and Astrophysics , 2019, Publications of the Astronomical Society of the Pacific.

[5]  C. A. Oxborrow,et al.  Planck2018 results , 2018, Astronomy & Astrophysics.

[6]  Dominik Bomans,et al.  Sharpening up Galactic all-sky maps with complementary data , 2018, Astronomy & Astrophysics.

[7]  Torsten A. Enßlin,et al.  Information Theory for Fields , 2018, Annalen der Physik.

[8]  Luciana Bianchi,et al.  Revised Catalog of GALEX Ultraviolet Sources. I. The All-Sky Survey: GUVcat_AIS , 2017, 1704.05903.

[9]  Christian Igel,et al.  Big Universe, Big Data: Machine Learning and Image Analysis for Astronomy , 2017, IEEE Intelligent Systems.

[10]  Torsten A. Enßlin,et al.  Optimal Belief Approximation , 2016, Entropy.

[11]  B. Winkel,et al.  HI4PI: a full-sky H i survey based on EBHIS and GASS , 2016, 1610.06175.

[12]  David M. Blei,et al.  Variational Inference: A Review for Statisticians , 2016, ArXiv.

[13]  G. W. Pratt,et al.  Planck 2015 results - X. Diffuse component separation: Foreground maps , 2015, 1502.01588.

[14]  M. Remazeilles,et al.  An improved source-subtracted and destriped 408 MHz all-sky map , 2014, 1411.3628.

[15]  T. Ensslin,et al.  The denoised, deconvolved, and decomposed Fermi γ-ray sky - An application of the D3PO algorithm , 2014, 1410.4562.

[16]  C. A. Oxborrow,et al.  Planck 2015 results. I. Overview of products and scientific results , 2015 .

[17]  Vlasios Vasileiou,et al.  THE FERMI LARGE AREA TELESCOPE ON ORBIT: EVENT CLASSIFICATION, INSTRUMENT RESPONSE FUNCTIONS, AND CALIBRATION , 2012 .

[18]  T. Nakagawa,et al.  The AKARI far-infrared all-sky survey maps , 2012, 1503.06421.

[19]  Pascal Vincent,et al.  Representation Learning: A Review and New Perspectives , 2012, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[20]  N. Evans,et al.  Star Formation in the Milky Way and Nearby Galaxies , 2012, 1204.3552.

[21]  Christopher N. Beaumont,et al.  CLASSIFYING STRUCTURES IN THE INTERSTELLAR MEDIUM WITH SUPPORT VECTOR MACHINES: THE G16.05-0.57 SUPERNOVA REMNANT , 2011, 1107.5584.

[22]  J. Chiang,et al.  THE LARGE AREA TELESCOPE ON THE FERMI GAMMA-RAY SPACE TELESCOPE MISSION , 2009, 0902.1089.

[23]  Jean-François Cardoso,et al.  Component Separation With Flexible Models—Application to Multichannel Astrophysical Observations , 2008, IEEE Journal of Selected Topics in Signal Processing.

[24]  R. B. Barreiro,et al.  Component separation methods for the PLANCK mission , 2008, 0805.0269.

[25]  H. K. Eriksen,et al.  Joint Bayesian Component Separation and CMB Power Spectrum Estimation , 2007, 0709.1058.

[26]  Nasser M. Nasrabadi,et al.  Pattern Recognition and Machine Learning , 2006, Technometrics.

[27]  Geoffrey E. Hinton,et al.  Reducing the Dimensionality of Data with Neural Networks , 2006, Science.

[28]  Donald P. Cox,et al.  THE THREE-PHASE INTERSTELLAR MEDIUM REVISITED , 2005 .

[29]  G. Lagache,et al.  IRIS: A New Generation of IRAS Maps , 2004, astro-ph/0412216.

[30]  K. Gorski,et al.  HEALPix: A Framework for High-Resolution Discretization and Fast Analysis of Data Distributed on the Sphere , 2004, astro-ph/0409513.

[31]  Douglas P. Finkbeiner,et al.  A Full-Sky Hα Template for Microwave Foreground Prediction , 2003, astro-ph/0301558.

[32]  J. Cardoso,et al.  Multidetector multicomponent spectral matching and applications for cosmic microwave background data analysis , 2002, astro-ph/0211504.

[33]  P. Reich,et al.  A radio continuum survey of the southern sky at 1420 MHz - The atlas of contour maps , 2001 .

[34]  D. Van Buren,et al.  A Robotic Wide‐Angle Hα Survey of the Southern Sky , 2001, astro-ph/0108518.

[35]  K. Ferrière The interstellar environment of our galaxy , 2001, astro-ph/0106359.

[36]  L. Haffner The Wisconsin H-Alpha Mapper Northern Sky Survey of Galactic Ionized Hydrogen , 2000, astro-ph/0112232.

[37]  Dan McCammon,et al.  ROSAT Survey Diffuse X-Ray Background Maps. II. , 1997 .

[38]  Dan McCammon,et al.  First Maps of the Soft X-Ray Diffuse Background from the ROSAT XRT/PSPC All-Sky Survey , 1995 .

[39]  Dan McCammon,et al.  A model for the distribution of material generating the soft X-ray background , 1990 .

[40]  Geoffrey E. Hinton,et al.  Learning internal representations by error propagation , 1986 .

[41]  de T. Jong,et al.  The Infrared Astronomical Satellite (IRAS) mission , 1984 .

[42]  E. Jaynes On the rationale of maximum-entropy methods , 1982, Proceedings of the IEEE.

[43]  J. Nousek,et al.  Soft diffuse X-rays in the southern galactic hemisphere , 1977 .

[44]  E. Purcell,et al.  Observation of a Line in the Galactic Radio Spectrum: Radiation from Galactic Hydrogen at 1,420 Mc./sec. , 1951, Nature.

[45]  M. Freyberg New ROSAT PSPC all-sky survey maps of the diffuse soft X-ray background , 1998 .

[46]  F. Kahn Hot Plasmas in Supernova Remnants , 1980 .

[47]  V. L. Ginzburg,et al.  Cosmic Magnetobremsstrahlung (Synchrotron Radiation) , 1965 .