Parameter estimation from a model grid application to the Gaia RVS spectra

In the framework of the ESA Gaia mission, stellar atmospheric parameters will be extracted for millions of spectra that will be observed by Gaia RVS (Wilkinson et al. 2005) [21]. Due to this high number of observed spectra it is necessary that the analysis be carried out using fast and robust automated algorithms. In this paper, we analyze the efficiency of a selection of fitting algorithms in obtaining stellar parameters for a sample of spectra. Several of these algorithms are based on the use of a decision tree, either oblique, kd or decorated. The tests are carried out using the same model grid in the same software environment. Different performance indices associated with our scientific goal are examined. The application of the Gauss–Newton algorithm initialized using a decision tree algorithm appeared to best satisfy the performance criteria.