Machine learning the gap between real and simulated nebulae: A domain-adaptation approach to classify ionised nebulae in nearby galaxies

Abstract

Classifying ionised nebulae in nearby galaxies is crucial to studying stellar feedback mechanisms and understanding the physical conditions of the interstellar medium. This classification task is generally performed by comparing observed line ratios with photoionisation simulations of different types of nebulae (HII regions, planetary nebulae, and supernova remnants). However, due to simplifying assumptions, such simulations are generally unable to fully reproduce the line ratios in observed nebulae. This discrepancy limits the performance of the classical machine-learning approach, where a model is trained on the simulated data and then used to classify real nebulae. For this study, we used a domain-adversarial neural network (DANN) to bridge the gap between photoionisation models (source domain) and observed ionised nebulae from the PHANGS-MUSE survey (target domain). The DANN is an example of a domain-adaptation algorithm, whose goal is to maximise the performance of a model trained on labelled data in the source domain on an unlabelled target domain by extracting domain-invariant features. Our results indicate a significant improvement in classification performance in the target domain when employing the DANN framework compared to a classical neural network (NN) classifier. Additionally, we investigated the impact of adding noise to the source dataset, finding that noise injection acts as a form of regularisation, further enhancing the performances of both the NN and DANN models on the observational data. The combined use of domain adaptation and noise injection improved the classification accuracy in the target domain by 23%. This study highlights the potential of domain adaptation methods in tackling the domain-shift challenge when using theoretical models to train machine-learning pipelines in astronomy.

Accepted in A&A

References in corpus (44)

• The Host Galaxies of AGN
• Theoretical Modeling of Starburst Galaxies
• The Host Galaxies and Classification of Active Galactic Nuclei
• The MAPPINGS III Library of Fast Radiative Shock Models
• De Re Metallica: The cosmic chemical evolution of galaxies
• A comprehensive classification of galaxies in the SDSS: How to tell true from fake AGN?
• The PHANGS-MUSE survey -- Probing the chemo-dynamical evolution of disc galaxies
• SDSS-IV MaNGA: The Impact of Diffuse Ionized Gas on Emission-line Ratios, Interpretation of Diagnostic Diagrams, and Gas Metallicity Measurements
• Dispersal of molecular clouds by ionising radiation
• Deriving model-based T${}_e$-consistent chemical abundances in ionised gaseous nebulae
• The Role of Stellar Feedback in the Dynamics of HII Regions
• Ionization Correction Factors for Planetary Nebulae: I- Using optical spectra
• The MAGNUM survey: different gas properties in the outflowing and disk components in nearby active galaxies with MUSE
• Nebular spectroscopy: A guide on H II regions and planetary nebulae
• A tale of two DIGs: The relative role of HII regions and low-mass hot evolved stars in powering the diffuse ionised gas (DIG) in PHANGS-MUSE galaxies
• Deep learning predictions of galaxy merger stage and the importance of observational realism
• Theoretical ISM pressure and electron density diagnostics for local and high-redshift galaxies
• Smoothed Particle Hydrodynamics simulations of expanding HII regions. I. Numerical methods and tests
• Identifying Galaxy Mergers in Observations and Simulations with Deep Learning
• Dimensionality Reduction of SDSS Spectra with Variational Autoencoders
• Which feedback mechanisms dominate in the high-pressure environment of the Central Molecular Zone?
• SIGNALS: I. Survey Description
• Self-consistent predictions for LIER-like emission lines from post-AGB stars
• Temperature inhomogeneities cause the abundance discrepancy in H II regions
• The metallicity's fundamental dependence on both local and global galactic quantities
• Ionization correction factors and dust depletion patterns in giant HII regions
• A diagnostic tool for the identification of Supernova Remnants
• Density biases and temperature relations for DESIRED HII regions
• The impact of pre-supernova feedback and its dependence on environment
• Discrepancies between the [OIII] and [SIII] Temperatures in HII Regions
• DeepMerge II: Building Robust Deep Learning Algorithms for Merging Galaxy Identification Across Domains
• MUSE crowded field 3D spectroscopy in NGC300 I. First results from central fields
• WARPFIELD-EMP: The Self-Consistent Prediction of Emission Lines from Evolving HII Regions in Dense Molecular Clouds
• Galaxy gas flows inferred from a detailed, spatially resolved metal budget
• PHANGS-MUSE: Detection and Bayesian classification of ~40000 ionised nebulae in nearby spiral galaxies
• Planetary Nebula Luminosity Function distances for 19 galaxies observed by PHANGS-MUSE
• WARPFIELD Population Synthesis: The physics of (extra-)Galactic star formation and feedback driven cloud structure and emission from sub-to-kpc scales
• Capturing the physics of MaNGA galaxies with self-supervised Machine Learning
• Theoretically Modelling Photoionized Regions with Fractal Geometry in Three Dimensions
• Emission-line diagnostics of HII regions using conditional Invertible Neural Networks
• Discovery of $\sim$2200 new supernova remnants in 19 nearby star-forming galaxies with MUSE spectroscopy
• MUSE crowded field 3D spectroscopy in NGC 300 III. Characterizing extremely faint HII regions and diffuse ionized gas
• Noise-Net: Determining physical properties of HII regions reflecting observational uncertainties
• Inferring redshift and galaxy properties via a multi-task neural net with probabilistic outputs: An application to simulated MOONS spectra

Cited by in corpus (5)

• Confirmation of the planetary nebula nature of HaTr 5. Not the remnant of Nova Sco 1437
• Automated quasar continuum estimation using neural networks: a comparative study of deep-learning architectures
• Classifying spectra of emission-line regions with neural networks -- An application to integral field spectroscopic data of M33
• LenNet: Direct Detection and Localization of Strong Gravitational Lenses in Wide-Field Sky Survey Images
• Deep MUSE observations of SNR 0509-67.5 reveal a double degenerate merger progenitor