Morphological Galaxy Classification with Shapelets

We present an unsupervised classification algorithm, that identifies natural classes of galaxy morphologies. Working on SDSS G‐band imaging data, we encode the morphologies by shapelet decomposition. The algorithm employs a model‐based soft clustering analysis to find groupings of similar data points. We demonstrate that the algorithm is able to clearly identify and distinguish groups of elliptical, face‐on and edge‐on spiral galaxies in a training data set. Based on the soft clustering results, we set up a soft classifier for a data set containing 1602 SDSS galaxies.