Classification of Stellar Spectra with LLE

TL;DR

This paper demonstrates that local linear embedding effectively reduces the dimensionality of stellar spectra, enabling accurate hierarchical classification aligned with traditional spectral types without feature extraction.

Contribution

It introduces a novel hierarchical classification method using LLE that accurately classifies stellar spectra and reproduces MK classifications without feature extraction.

Findings

Most stellar spectra form a one-dimensional sequence in 3D space.

Position along the sequence correlates with spectral temperature.

The method can identify spectra with emission or broad absorption lines.

Abstract

We investigate the use of dimensionality reduction techniques for the classification of stellar spectra selected from the SDSS. Using local linear embedding (LLE), a technique that preserves the local (and possibly non-linear) structure within high dimensional data sets, we show that the majority of stellar spectra can be represented as a one dimensional sequence within a three dimensional space. The position along this sequence is highly correlated with spectral temperature. Deviations from this "stellar locus" are indicative of spectra with strong emission lines (including misclassified galaxies) or broad absorption lines (e.g. Carbon stars). Based on this analysis, we propose a hierarchical classification scheme using LLE that progressively identifies and classifies stellar spectra in a manner that requires no feature extraction and that can reproduce the classic MK classifications to an accuracy of one type.