ASCL.net

Astrophysics Source Code Library

Making codes discoverable since 1999

ASCL Code Record

[ascl:2502.017] blasé: Interpretable machine learning for high-resolution astronomical spectroscopy

blasé performs whole-spectrum fitting by cloning 10,000+ spectral lines from a pre-computed synthetic spectral model template and then learning the perturbations to those lines through comparison to real data. Each spectral line has four parameters, yielding possibly 40,000+ parameters. The technique uses autodiff to tune the parameters precisely and quickly. Built in PyTorch with native GPU support, blasé can be extended to, for example, Doppler imaging, Precision RVs, and abundances.

Code site:
https://github.com/gully/blase
Described in:
https://ui.adsabs.harvard.edu/abs/2022ApJ...941..200G
Bibcode:
2025ascl.soft02017G

Views: 12

ascl:2502.017
Add this shield to your page
Copy the above HTML to add this shield to your code's website.