Realising the potential of large spectroscopic surveys with machine-learning (2503.08196v1)

Abstract: Machine-learning is playing an increasing role in helping the astronomical community to face data analysis challenges, in particular in the field of Galactic Archaeology and large scale spectroscopic surveys. We present recent developments in the field of convolutional neural-networks (CNNs) for stellar abundances in the context of the Galactic spectroscopic surveys Gaia-ESO, and Gaia-RVS. Especially, by combining the full Gaia data product, we manage to characterize for the first time the [alpha/M] vs. [M/H] bimodality in the Galactic disc with Gaia-RVS spectra at low-S/N. This work is highly relevant for the next generation of large scale surveys such as MSE, 4MOST, and WST.