Estimating stellar atmospheric parameters based on LASSO and support-vector regression

TL;DR

This paper presents a novel method combining Haar wavelet, LASSO, and support-vector regression to accurately estimate stellar atmospheric parameters from spectral data, validated on multiple datasets.

Contribution

The study introduces a new scheme integrating wavelet decomposition, feature selection, and regression for stellar parameter estimation, improving accuracy over existing methods.

Findings

Achieved low mean absolute errors on SDSS, LAMOST, and synthetic spectra datasets.

Demonstrated effectiveness of combined wavelet, LASSO, and SVR approach.

Validated the method's robustness across different spectral datasets.

Abstract

A scheme for estimating atmospheric parameters T$_{eff}$, log$~g$, and [Fe/H] is proposed on the basis of Least Absolute Shrinkage and Selection Operator (LASSO) algorithm and Haar wavelet. The proposed scheme consists of three processes. A spectrum is decomposed using the Haar wavelet transform and low-frequency components at the fourth level are considered as candidate features. Then, spectral features from the candidate features are detected using the LASSO algorithm to estimate the atmospheric parameters. Finally, atmospheric parameters are estimated from the extracted spectral features using the support-vector regression (SVR) method. The proposed scheme was evaluated using three sets of stellar spectra respectively from Sloan Digital Sky Survey (SDSS), Large Sky Area Multi-object Fiber Spectroscopic Telescope (LAMOST), and Kurucz's model, respectively. The mean absolute errors are as follows: for 40~000 SDSS spectra, 0.0062 dex for log~T$_{eff}$ (85.83 K for T$_{eff}$), 0.2035 dex for log$~g$ and 0.1512 dex for [Fe/H]; for 23963 LAMOST spectra, 0.0074 dex for log~T$_{eff}$ (95.37 K for T$_{eff}$), 0.1528 dex for log~$g$, and 0.1146 dex for [Fe/H]; and for 10469 synthetic spectra, 0.0010 dex for log T$_{eff}$(14.42K for T$_{eff}$), 0.0123 dex for log~$g$, and 0.0125 dex for [Fe/H].