Recovering buoyancy-glitch parameters in low-luminosity red giants from global stellar properties with machine learning (code, catalogue and model weights)

Training scripts, experiment configurations, the reconstructed buoyancy-glitch-parameter catalogue (12,580 ASTEC/ADIPLS stellar models), the fixed train/validation/test splits, and the trained gradient-boosted tree regressor (XGBoost/LightGBM) model weights accompanying the MNRAS submission 'Recovering buoyancy-glitch parameters in low-luminosity red giants from global stellar properties with machine learning'. The glitch targets are posterior medians of the finite-width buoyancy-glitch prescription of Cunha et al. (2019) fitted to the local buoyancy-frequency profile of each model.