Deep learning inference with the Event Horizon Telescope II. The Zingularity framework for Bayesian artificial neural networks

Abstract

(abridged) In this second paper in our publication series, we present the open-source Zingularity framework for parameter inference with deep Bayesian artificial neural networks. We carried out out supervised learning with synthetic millimeter very long baseline interferometry observations of the EHT. Our ground-truth models are based on GRMHD simulations of Sgr A* and M87* on horizon scales. We investigated how well Zingularity neural networks are able to infer key model parameters from EHT observations, such as the black hole spin and the magnetic state of the accretion disk, when uncertainties in the data are accurately taken into account. Zingularity makes use of the TensorFlow Probability library and is able to handle large amounts of data with a combination of the efficient TFRecord data format plus the Horovod framework. Our approach is the first analysis of EHT data with Bayesian neural networks, where an unprecedented training data size, under consideration of a closely modeled EHT signal path, and the full information content of the observational data are used. Zingularity infers parameters based on salient features in the data and is containerized. Through parameter surveys and dedicated validation tests, we identified neural network architectures, that are robust against internal stochastic processes and unaffected by noise in the observational and model data. We give examples of how different data properties affect the network training. We show how the Bayesian nature of our networks gives trustworthy uncertainties and uncovers failure modes for uncharacterizable data. It is easy to achieve low validation errors during training on synthetic data with neural networks, particularly when the forward modeling is too simplified. Through careful studies, we demonstrate that our trained networks can generalize well so that reliable results can be obtained from observational data.

19 pages, 6 figures, A&A submitted 16/01/2025, accepted 30/03/2025, published 07/06/2025

References in corpus (113)

• First M87 Event Horizon Telescope Results. VI. The Shadow and Mass of the Central Black Hole
• Reconciling modern machine learning practice and the bias-variance trade-off
• First M87 Event Horizon Telescope Results. IV. Imaging the Central Supermassive Black Hole
• First M87 Event Horizon Telescope Results. V. Physical Origin of the Asymmetric Ring
• First M87 Event Horizon Telescope Results. II. Array and Instrumentation
• Detection of the gravitational redshift in the orbit of the star S2 near the Galactic centre massive black hole
• General Relativistic Magnetohydrodynamic Simulations of Magnetically Choked Accretion Flows around Black Holes
• First M87 Event Horizon Telescope Results. III. Data Processing and Calibration
• A geometric distance measurement to the Galactic Center black hole with 0.3% uncertainty
• First M87 Event Horizon Telescope Results. VIII. Magnetic Field Structure near The Event Horizon
• First Sagittarius A* Event Horizon Telescope Results. VI: Testing the Black Hole Metric
• General Relativistic Magnetohydrodynamic Simulations of Jet Formation and Large-Scale Propagation from Black Hole Accretion Systems
• Detection of the Schwarzschild precession in the orbit of the star S2 near the Galactic centre massive black hole
• First Sagittarius A* Event Horizon Telescope Results. V. Testing Astrophysical Models of the Galactic Center Black Hole
• GRMHD Simulations of Magnetized Advection Dominated Accretion on a Non-Spinning Black Hole: Role of Outflows
• Relativistic redshift of the star S0-2 orbiting the Galactic center supermassive black hole
• The Black-Hole Mass in M87 from Gemini/NIFS Adaptive Optics Observations
• First Sagittarius A* Event Horizon Telescope Results. III: Imaging of the Galactic Center Supermassive Black Hole
• Global General Relativistic MHD Simulation of a Tilted Black-Hole Accretion Disk
• First Sagittarius A* Event Horizon Telescope Results. II. EHT and Multi-wavelength Observations, Data Processing, and Calibration
• Universal Interferometric Signatures of a Black Hole's Photon Ring
• First Sagittarius A* Event Horizon Telescope Results. IV. Variability, Morphology, and Black Hole Mass
• Constraints on black-hole charges with the 2017 EHT observations of M87*
• Gravitational Test Beyond the First Post-Newtonian Order with the Shadow of the M87 Black Hole
• The Structure and Dynamics of the Sub-parsec Scale Jet in M87 Based on 50 VLBA Observations Over 17 Years at 43 GHz
• The mass distribution in the Galactic Centre from interferometric astrometry of multiple stellar orbits
• Introduction to astroML: Machine Learning for Astrophysics
• The Black Hole Accretion Code
• Black hole flares: ejection of accreted magnetic flux through 3D plasmoid-mediated reconnection
• Magnetic Reconnection and Hot Spot Formation in Black Hole Accretion Disks
• Kinematics of the jet in M87 on scales of 100 -- 1000 Schwarzschild radii
• A High Resolution X-ray Image of the Jet in M 87
• The Shadow of a Spherically Accreting Black Hole
• Alignment of Magnetized Accretion Disks and Relativistic Jets with Spinning Black Holes
• Observing the Inner Shadow of a Black Hole: A Direct View of the Event Horizon
• Semi-implicit scheme for treating radiation under M1 closure in general relativistic conservative fluid dynamics codes
• Two-temperature, Magnetically Arrested Disc simulations of the jet from the supermassive black hole in M87
• The size of the jet launching region in M87
• Bardeen-Petterson Alignment, Jets and Magnetic Truncation in GRMHD Simulations of Tilted Thin Accretion Discs
• The role of electron heating physics in images and variability of the Galactic Centre black hole Sagittarius A*
• Radiative, two-temperature simulations of low luminosity black hole accretion flows in general relativity
• Surveying the reach and maturity of machine learning and artificial intelligence in astronomy
• Key Science Goals for the Next-Generation Event Horizon Telescope
• Can the EHT M87 results be used to test general relativity?
• Hydrodynamic Simulations of the Inner Accretion Flow of Sagittarius A* Fueled By Stellar Winds
• Millimeter light curves of Sagittarius A* observed during the 2017 Event Horizon Telescope campaign
• Broadband Multi-wavelength Properties of M87 during the 2017 Event Horizon Telescope Campaign
• Polarimetric properties of Event Horizon Telescope targets from ALMA
• Global General Relativistic Magnetohydrodynamic Simulations of Accretion Tori
• The DAWES review 10: The impact of deep learning for the analysis of galaxy surveys
• Pair Production in Low Luminosity Galactic Nuclei
• Images and photon ring signatures of thick disks around black holes
• Photon rings of spherically symmetric black holes and robust tests of non-Kerr metrics
• Modeling non-thermal emission from the jet-launching region of M 87 with adaptive mesh refinement
• First M87 Event Horizon Telescope Results. IX. Detection of Near-horizon Circular Polarization
• General relativistic magnetohydrodynamical $\kappa$-jet models for Sgr A*
• Inflow-Outflow Model with Conduction and Self-Consistent Feeding for Sgr A*
• State-of-the-art energetic and morphological modelling of the launching site of the M87 jet
• Multi-dimensional simulations of ergospheric pair discharges around black holes
• Two-Temperature GRRMHD Simulations of M87
• GRMHD simulations of accretion onto Sgr A*: How important are radiative losses?
• Discriminating Accretion States via Rotational Symmetry in Simulated Polarimetric Images of M87
• RAPTOR I: Time-dependent radiative transfer in arbitrary spacetimes
• Reference Array and Design Consideration for the next-generation Event Horizon Telescope
• Rapid Variability of Sgr A* across the Electromagnetic Spectrum
• Astronomia ex machina: a history, primer, and outlook on neural networks in astronomy
• Data-Driven Reconstruction of Gravitationally Lensed Galaxies using Recurrent Inference Machines
• Closure statistics in interferometric data
• Comparison of the ion-to-electron temperature ratio prescription: GRMHD simulations with electron thermodynamics
• Spectral evolution of flaring blazars from numerical simulations
• Photon ring test of the Kerr hypothesis: Variation in the ring shape
• General relativistic MHD simulations of non-thermal flaring in Sagittarius A*
• Relativistic resistive magnetohydrodynamic reconnection and plasmoid formation in merging flux tubes
• The Image of the M87 Black Hole Reconstructed with PRIMO
• Tilted Disks around Black Holes: A Numerical Parameter Survey for Spin and Inclination Angle
• Iharm3D: Vectorized General Relativistic Magnetohydrodynamics
• Conservative GRMHD Simulations of Moderately Thin, Tilted Accretion Disks
• Black Hole Images as Tests of General Relativity: Effects of Plasma Physics
• Particle acceleration in kink-unstable jets
• Spectral and Imaging properties of Sgr A* from High-Resolution 3D GRMHD Simulations with Radiative Cooling
• Black Hole Polarimetry I: A Signature of Electromagnetic Energy Extraction
• RAPTOR II: Polarized radiative transfer in curved spacetime
• Images of magnetospheric reconnection-powered radiation around supermassive black holes
• Deep Learning-based Imaging in Radio Interferometry
• On the Comparison of AGN with GRMHD Simulations: I. Sgr A*
• Machine Learning for Observational Cosmology
• Unraveling Twisty Linear Polarization Morphologies in Black Hole Images
• Deep Horizon; a machine learning network that recovers accreting black hole parameters
• Using evolutionary algorithms to model relativistic jets: Application to NGC 1052
• Using Machine Learning to Link Black Hole Accretion Flows with Spatially Resolved Polarimetric Observables
• Bayesian techniques for comparing time-dependent GRMHD simulations to variable Event Horizon Telescope observations
• Closure Traces: Novel Calibration-Insensitive Quantities for Radio Astronomy
• A D-term Modeling Code (DMC) for simultaneous calibration and full-Stokes imaging of very long baseline interferometric data
• An Updated Formalism For Line-Driven Radiative Acceleration and Implications for Stellar Mass Loss
• Bayesian Accretion Modeling: Axisymmetric Equatorial Emission in the Kerr Spacetime
• Rotation in Event Horizon Telescope Movies
• Radiation GRMHD simulations of M87: funnel properties and prospects for gap acceleration
• General relativistic hydrodynamic simulations of perturbed transonic accretion
• Using multiobjective optimization to reconstruct interferometric data (I)
• alpha-Deep Probabilistic Inference (alpha-DPI): efficient uncertainty quantification from exoplanet astrometry to black hole feature extraction
• The Variability of the Black-Hole Image in M87 at the Dynamical Time Scale
• Principal-Component Interferometric Modeling (PRIMO), an Algorithm for EHT Data I: Reconstructing Images from Simulated EHT Observations
• Effects of Hydrogen vs. Helium on Electromagnetic Black Hole Observables
• Event-horizon-scale Imaging of M87* under Different Assumptions via Deep Generative Image Priors
• Swarm intelligence for full Stokes dynamic imaging reconstruction of interferometric data
• Black Hole Weather Forecasting with Deep Learning: A Pilot Study
• Deep learning inference with the Event Horizon Telescope III. Zingularity results from the 2017 observations and predictions for future array expansions
• Deep learning inference with the Event Horizon Telescope I. Calibration improvements and a comprehensive synthetic data library
• A probabilistic approach to phase calibration: I. Effects of source structure on fringe-fitting
• Neural Network Astronomy as a New Tool for Observing Bright and Compact Objects
• Interferometric Image Reconstruction using Closure Invariants and Machine Learning
• A proof-of-concept neural network for inferring parameters of a black hole from partial interferometric images of its shadow
• Impact of the nonthermal electron radiation effects on the horizon scale image structure of Sagittarius A*

Cited by in corpus (3)

• Deep learning inference with the Event Horizon Telescope III. Zingularity results from the 2017 observations and predictions for future array expansions
• Deep learning inference with the Event Horizon Telescope I. Calibration improvements and a comprehensive synthetic data library
• Observational properties of black hole in quantum fluctuation modified gravity