Reliability checks on the Indo-US Stellar Spectral Library using Artificial Neural Networks and Principal Component Analysis

Abstract

The Indo-US coudé feed stellar spectral library (CFLIB) made available to the astronomical community recently by Valdes et al. (2004) contains spectra of 1273 stars in the spectral region 3460 to 9464 Åat a high resolution of 1 ÅFWHM and a wide range of spectral types. Cross-checking the reliability of this database is an important and desirable exercise since a number of stars in this database have no known spectral types and a considerable fraction of stars has not so complete coverage in the full wavelength region of 3460-9464 Å resulting in gaps ranging from a few Åto several tens of Å. In this paper, we use an automated classification scheme based on Artificial Neural Networks (ANN) to classify all 1273 stars in the database. In addition, principal component analysis (PCA) is carried out to reduce the dimensionality of the data set before the spectra are classified by the ANN. Most importantly, we have successfully demonstrated employment of a variation of the PCA technique to restore the missing data in a sample of 300 stars out of the CFLIB.