Unresolved Galaxy Classifier for ESA's Gaia Mission
I. Bellas-Velidis, Mary Kontizas, E. Livanou, Paraskevi Tsalmantza · ASPC · 2010
Unresolved Galaxy Classifier (UGC), a software package for the groundbased pipeline of ESA’s Gaia mission is presented. It aims at analyzing Gaia BP/RP spectra of unresolved galaxies, to provide taxonomic classification and specific parameters estimation. The UGC algorithm is based on Support Vector Machines, a supervised learning technique. The software is implemented in JAVA. An offline UGC-learning module provides functions for SVM-model training. Once trained, the set of models can be repeatedly applied to unknown galaxy spectra by the pipeline’s UGC-application module. Tests with a library of BP/RP simulated galaxy spectra show a very good performance of UGC.