Classification of Astronomical Spectra Using PCA-Compressed Flux and Inverse-Variance Features
Researchers have developed a new pipeline for classifying astronomical spectra, utilizing Principal Component Analysis (PCA) for feature compression and the LightGBM classifier for improved accuracy. This method represents each spectrum by its flux and inverse-variance information, which is then compressed and concatenated. The LightGBM model achieved a notable 94.6% accuracy and 92.1% balanced accuracy in distinguishing between stars, galaxies, and quasars from the SDSS DR17 dataset. AI
IMPACT This research demonstrates an effective application of machine learning for scientific data analysis, potentially improving the efficiency of astronomical surveys.