E(B-V) DETERMINATIONS OF O-STAR AND B-STAR USING ARTIFICIAL NEURAL NETWORKS

We propose a new method to determine E(B - V) from low-dispersion ultr aviolet spectra by using artificial neural networks. This method has b een applied to O and B stars in the IUE low-dispersion catalog. A comp arison of the ANN-based determinations with the classical method of co mparing observed (B - V) with intrinsic color (B - V)(0) is performed to assess the reliability and limitations of the new method. We have b een able to determine E(B - V) to an accuracy of 0.08 mag and also bee n able to single out a star which follows an anomalous interstellar ex tinction law.