CLUES: Interpretable Unsupervised Clustering of Debris-Disk Spectra
A non-parametric, fully-interpretable machine-learning pipeline (CLUES - CLustering UnsupErvised with Sequencer) that classifies debris-disk infrared spectra by combining several unsupervised clustering algorithms with multi-scale distance measures, including minimum-spanning-tree/Sequencer ordering. Applied to the Spitzer IRS debris-disk catalog, it discovers new compositional groupings and mineralogical trends in silicate-emission spectra without a priori spectral templates, supporting systematic study of debris-disk mineralogy during terrestrial planet formation.
Draft version 2025/01/06 Typeset using LATEX twocolumn style in AASTeX631 Sequencing Silicates in
Introduces CLUES (CLustering UnsupErvised with Sequencer), a non-parametric, fully-interpretable machine-learning tool for spectral analysis of debris disks. CLUES combines multiple unsupervised clus…