Machine Learning (ML) is reshaping the way we process and analyse galactic and extragalactic data. While much of the early progress has focused on imaging, the systematic application of ML to spectroscopy is rapidly emerging as a transformative frontier. Spectra encode the richest information on galaxies and AGN, from stellar populations and chemical enrichment to black hole masses and feedback signatures. Their growing volume makes ML essential to unlock deeper insights from these data, especially in the era of existing and upcoming large spectroscopic surveys enabled by instruments like SDSS, DESI, MOONS, Euclid, JWST, and the forthcoming ELT, and enriched by complementary data from the X-ray (e.g. Chandra, XMM-Newton, eROSITA) and radio bands (e.g. LOFAR, SKA pathfinders).

Novel ML methods now allow for full-spectrum analysis across all wavelengths without reliance on pre-defined templates, robust extraction of weak or blended features, and automated identification of rare or unexpected sources. Neural networks can efficiently compress spectral information and reveal previously hidden patterns. Also, by leveraging synthetic data generated with state-of-the-art forward modelling tools, simulation-based inference with neural flows can closely reproduce the behaviour of Bayesian frameworks. Yet these examples represent only a glimpse of the broader transformative potential of ML in this domain.

This session will highlight cutting-edge applications of ML to galactic and extragalactic spectroscopy, emphasising current and expected scientific breakthroughs. By connecting diverse research communities, we will showcase how ML is accelerating a deeper understanding of galaxies and AGN physics, setting the stage for next-generation spectroscopic surveys.

Programme

We will bring together researchers working on galaxies, AGN, and survey design to address the following key questions:

• What major scientific advances in galaxy and AGN studies have been enabled by applying ML to spectroscopy, and how do we envision the next steps in this field?
• How can ML tools effectively enhance spectral data processing, redshift determination, physical parameter inference, and optimisation of survey strategies?
• How should we tackle key challenges such as uncertainties, interpretability, and the transfer of knowledge across surveys, instruments, and domains?

Invited speakers

• Chris Lovell (Kavli Institute for Cosmology, Cambridge, UK)
• Małgorzata Siudek (Institute of Astrophysics of the Canary Islands (IAC, Tenerife)
• Francesco Belfiore (ESO-Garching, Germany)

Scientific organisers

Susanna Bisogni - INAF-IASF Milano (Italy, co-chair)
Adriana Gargiulo - INAF-IASF Milano (Italy, co-chair)
Michele Ginolfi - Università degli Studi di Firenze, INAF-Arcetri Astrophysical Observatory (Italy, co-chair)
Caterina Bracci - Università degli studi di Firenze (Italy)
Patricia Iglesia Navarro - Instituto de Astrofísica de Canarias (Spain)
Ralf Klessen - Universität Heidelberg, Zentrum für Astronomie Institut für Theoretische Astrophysik (Germany)
Vivienne Wild - School of Physics and Astronomy, University of St. Andrews (UK)
Fucheng Zhong - School of Physics and Astronomy, Sun Yat-sen University (P. R. China)

Contact

Susanna Bisogni: susanna.bisogni @ inaf.it,
Adriana Gargiulo: adriana.gargiulo @ inaf.it,
Michele Ginolfi: michele.ginolfi @ unifi.it