Katsuhiro MURATA
(Nagoya University)
I discuss machine learning method to predict galaxy morphologies without colors.
Previous automatic classifications via machine learning algorithm sometimes used their morphological features like CAS as well as their colors. Although it allows us to classify galaxy morphologies more precisely due to the color-morphology relation, it also introduces a potential bias of classification as a function of their colors: redder galaxies tend to be classified as earlier Hubble types. Therefore, machine learning with colors is not appropriate to the study of the relation between galaxy morphologies and other properties, such as colors, star formation rates or possibly AGN activities.
I developed a method using only morphological features of bright
SDSS galaxies to predict their morphologies with a machine learning algorithm based on support vector machine. I found that the machine predicts their morphologies as good as we can do with visual inspection.