Complete results of the SpectralAutoencoder applied to the 183 LAMOST DR5 features: symmetric MLP architecture, latent space z=2, MSE(AE)=0.489 ≈ PCA(k=10), gravity axis discovery on latent axis 1, QSO anomaly detection (7 QSOs, reconstruction error ≈ 4× the stellar median), interpolation and latent arithmetic.
Complete annotated list of the 183 spectroscopic features used in AstroSpectro (PHY-3500 x LAMOST DR5 x Gaia DR3), organised by physical family.
Complete description of the dataset used in the PHY-3500 project: N = 43,019 LAMOST DR5 spectra cross-matched with Gaia DR3, p = 183 spectroscopic features after quality filtering.
Complete comparison of three dimensionality reduction methods on 43,019 LAMOST DR5 stellar spectra: PCA (linear), UMAP (topological) and t-SNE (probabilistic). Hyperparameters, results, Procrustes stability and HDBSCAN clusters.
Complete guide to preprocessing LAMOST DR5 spectra in AstroSpectro: λ-grid reconstruction, median normalization, SNR/RUWE quality filters, and preparation of the 183-descriptor vector.
Complete presentation of LAMOST DR5 data used by AstroSpectro: telescope characteristics, FITS structure, SmartDownloader pipeline, Gaia DR3 cross-match and quality filters.
Detailed analysis of PCA principal axes applied to the 183 LAMOST DR5 features: explained variance, raw flux eigenspectra, loadings by spectral family, Gaia DR3 correlations and HR diagram.
AstroSpectro results in two separate studies: unsupervised dimensionality reduction (PCA, UMAP, t-SNE, HDBSCAN) on the PHY-3500 dataset (G1, 43,019 spectra), and the reference XGBoost classifier with its SHAP attribution on the 350k-spectrum generation (G2, run v6, which includes Gaia/photometric features).
UMAP topological projection and unsupervised HDBSCAN clustering of the stellar spectral feature space.
Results of the reference XGBoost classifier (run v6, 350k-spectrum generation G2, five types A–M): balanced accuracy, macro F1 and ROC-AUC from runs.json, per-class performance, SHAP attribution — with the Gaia/photometric features it contains — and its qualitative link with the PHY-3500 UMAP study (G1).