Automated Star‐Galaxy Discrimination for Large Surveys

Abstract

The size of survey data is increasing rapidly, and theautomatic classification of objects is becoming moreimportant. The classification is traditionally based, e.g.,on point- spread function (PSF) fitting. Recently, severaldifferent neural network approaches have been introducedfor classification. In this paper we use bothself-organized map and learning vector quantization basedneural networks for star-galaxy separation. Finally, wetest a hybrid algorithm using fuzzy classifier andback-propagation neural networks. We show that differentmethods give relatively similar results. The classificationaccuracy is good enough for real data analysis, andselection between different methods must be done based onalgorithmic complexity and availability of preclassifiedtraining sets.