I discuss various aspects of source classification and physical parametrization using data from the future Galactic survey mission GAIA. Due to the heterogeneity of the data, the large variety of objects observed and problems of data degeneracy (amongst other things), efficiently extracting physical information from these data will be challenging. I discuss the global and local nature of commonly used pattern recognition algorithms and outline two alternative frameworks for classification - parallel and hierarchical - and describe some aspects of each. A method for calibrating the classification algorithms is proposed which requires only a limited amount of additional (ground-based) data. By way of illustration, an example of stellar parametrization using GAIA-like RVS data is presented.
in GAIA Spectroscopy, Science and Technology, U. Munari (ed.),
ASP Conf. Ser. vol. 298, pp. 199-208
Astronomical Society of the
Pacific, San Francisco, in press
[PDF version] 10 pages, 261Kb