Stellar spectral classification based on principal component analysis and artificial neural networks

We demonstrate that the Principal Component Analysis (PCA), when used as a data pre-processing tool for the Neural Network Algorithms, provides an extremely efficient and robust procedure for classifying stellar spectra. We have been successful in classifying a library of optical stellar spectra with an accuracy similar to that obtained earlier, but at a fraction of the cpu time. It is hoped that this first step will allow us to handle and analyse large spectral databases planned for the future.