PhDLspec: physical-prior embedded deep learning method for spectroscopic determination of stellar labels in high-dimensional parameter space

Le résumé fourni par la source

PhDLspec is a tool for emulating stellar spectra and fitting stellar parameters in high-dimensional (>30) parameter space, using physical-prior embedded Transformer-based deep learning method. It trains a deep learning model on a pre-computed library of stellar spectra, and achieves precise emulation of both the spectral flux and its response functions to stellar parameters. Once trained, it can generate stellar spectra thousands of times faster than ab initio model computations. PhDLspec simultaneously infers ~30 parameters, including stellar atmospheric parameters (Teff, log g, [Fe/H]) and chemical elemental abundance ratios ([X/Fe]) from a single observed spectrum by rigorously utilizing their intrinsic physical features, employing either MCMC or CMA-ES optimization. If you use PhDLspec in your research, please cite: Wu, T. et al. (2026). PhDLspec: physical-prior embedded deep learning method for spectroscopic determination of stellar labels in high-dimensional parameter space.Zenodo. https://doi.org/10.5281/zenodo.19045278

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.