AN ARTIFICIAL NEURAL-NETWORK APPROACH TO THE CLASSIFICATION OF GALAXYSPECTRA

We present a method for the automated classification of galaxies with low signal-to-noise (S/N) ratio spectra typical of redshift surveys. W e develop spectral simulations based on the parameters for the 2-degre e-field Galaxy Redshift Survey, and with these simulations we investig ate the technique of principal component analysis when applied specifi cally to spectra of low S/N ratio. We relate the objective principal c omponents to features in the spectra and use a small number of compone nts to successfully reconstruct the underlying signal from the low-qua lity spectra. Using the principal components as input, we train an art ificial neural network to classify the noisy simulated spectra into mo rphological classes, revealing the success of the classification again st the observed b(J) magnitude of the source, which we compare with al ternative methods of classification. We find that more than 90 per cen t of our sample of normal galaxies are correctly classified into one o f the five broad morphological classes for simulations at b(J) = 19.7. By dividing the data into separate sets, we show that a classificatio n on to the Hubble sequence is relevant only for normal galaxies, and that spectra with unusual features should be incorporated into a class ification scheme based predominantly on their spectral signatures. We discuss how an artificial neural network can be used to distinguish no rmal and unusual galaxy spectra, and also discuss the possible applica tion of these results to spectra from galaxy redshift surveys.