Neural translation for Stokes inversion and synthesis

Abstract

[Abridged] The physical conditions in stellar atmospheres can be obtained from the interpretation of solar spectro-polarimetric observations. However, traditional inversion codes are computationally demanding, especially for lines whose formation is complex. The necessity of faster alternatives has motivated the emergence of machine learning solutions. This paper introduces an approach to the inversion and synthesis of Stokes profiles inspired by neural machine translation. Our aim is to develop a generative model that treats Stokes profiles and atmospheric models as two distinct ``languages'' encoding the same physical reality. We build a model that learns how to translate between them, also providing estimates of the uncertainty. We employ a tokenization strategy for both Stokes parameters and model atmospheres, which is learned using a VQ-VAE, a neural model used to compress the data into a lower dimensionality form. The core of our inversion code utilizes a transformer encoder-decoder architecture to perform the translation between these tokenized representations. The model is trained on a database of synthetic Stokes profiles derived from perturbations to various semi-empirical solar atmospheric models, ensuring a wide range of expected solar physical conditions. The method effectively reconstructs atmospheric models from observed Stokes profiles, showing better constrained models within the region of sensitivity of the considered spectral lines. The latent representation induced by the VQ-VAE helps accelerate the inversion by compressing the length of the Stokes profiles and model atmospheres. Additionally, it helps regularize the solution by reducing the chances of obtaining unphysical models. As a final advantage, the method provides the generative nature of our model, which naturally yields an estimate of the uncertainty in the solution.

14 pages, 9 figures, submitted to A&A

References in corpus (33)

• SciPy 1.0--Fundamental Algorithms for Scientific Computing in Python
• Array Programming with NumPy
• VFISV: Very Fast Inversion of the Stokes Vector for the Helioseismic and Magnetic Imager
• An Open Source, Massively Parallel Code for Non-LTE Synthesis and Inversion of Spectral Lines and Zeeman-induced Stokes Profiles
• Advanced Forward Modeling and Inversion of Stokes Profiles Resulting from the Joint Action of the Hanle and Zeeman Effects
• STiC -- A multi-atom non-LTE PRD inversion code for full-Stokes solar observations
• Inversion of the radiative transfer equation for polarized light
• Spatially coupled inversion of spectro-polarimetric image data I: Method and first results
• Radiative diagnostics in the solar photosphere and chromosphere
• Spectropolarimetric capabilities of Ca II 8542 A line
• Returning magnetic flux in sunspot penumbrae
• Spectropolarimetric NLTE inversion code SNAPI
• The usefulness of analytic response functions
• Stokes Inversion based on Convolutional Neural Networks
• A method for global inversion of multi-resolution solar data
• Comparison of inversion codes for polarized line formation in MHD simulations. I. Milne-Eddington codes
• DeSIRe: Departure coefficient aided Stokes Inversion based on Response functions
• Strategies for Spectral Profile Inversion using Artificial Neural Networks
• FIRTEZ-dz: A Forward and Inverse solver of the polarized Radiative Transfer Equation under Zeeman regime in geometrical scale
• GREGOR: Optics Redesign and Updates from 2018-2020
• Spatial deconvolution of spectropolarimetric data: an application to quiet Sun magnetic elements
• Analytical maximum likelihood estimation of stellar magnetic fields
• Machine learning initialization to accelerate Stokes profile inversions
• Zeeman-Tomography of the Solar Photosphere -- 3-Dimensional Surface Structures Retrieved from Hinode Observations
• The intrinsic dimensionality of spectro-polarimetric data
• Mimicking spectropolarimetric inversions using convolutional neural networks
• RADYNVERSION: Learning to Invert a Solar Flare Atmosphere with Invertible Neural Networks
• Feature Extraction Techniques for the Analysis of Spectral Polarization Profiles
• Exploring spectropolarimetric inversions using neural fields. Solar chromospheric magnetic field under the weak-field approximation
• Milne-Eddington inversion of the Fe I line pair at 630~nm
• Spectropolarimetric Inversion in Four Dimensions with Deep Learning (SPIn4D): I. Overview, Magnetohydrodynamic Modeling, and Stokes Profile Synthesis
• Convolutional Neural Networks and Stokes Response Functions
• Mapping the Sun's upper photosphere with artificial neural networks