3D Automated Classification Scheme for the TAUVEX data pipeline

Abstract

In order to develop a pipeline for automated classi cation of stars to be observed by the TAUVEX ultraviolet space Telescope, we employ an arti cial neural network (ANN) technique for classifying stars by using synthetic spectra in the UV region from 1250 A to 3220 A as the training set and International Ultraviolet Explorer (IUE) low resolution spectra as the test set. Both the data sets have been pre-processed to mimic the observations of the TAUVEX ultraviolet imager. We have successfully classi ed 229 stars from the IUE low resolution catalog to within 3-4 spectral sub-class using two di erent simulated training spectra, the TAUVEX spectra of 286 spectral types and UVBLUE spectra of 277 spectral types. Further, we have also been able to obtain the colour excess (i.e. E(B-V) in magnitude units) or the interstellar reddening for those IUE spectra which have known reddening to an accuracy of better than 0.1 magnitudes. It has been shown that even with the limitation of data from just photometric bands, ANNs have not only classi ed the stars, but also provided satisfactory estimates for interstellar extinction. The ANN based classi cation scheme has been successfully tested on the simulated TAUVEX data pipeline. It is expected that the same technique can be employed for data validation in the ultraviolet from the virtual observatories. Finally, the interstellar extinction estimated by applying the ANNs on the TAUVEX data base would provide an extensive extinction map for our galaxy and which could in turn be modeled for the dust distribution in the galaxy.