A robust morphological classification of high-redshift galaxies using support vector machines on seeing limited images. I Method description

TL;DR

This paper introduces a new automated support vector machine-based method for classifying high-redshift galaxies morphologically from seeing-limited images, achieving higher accuracy than classical methods and suitable for large surveys.

Contribution

The paper presents a novel non-parametric SVM approach for galaxy morphology classification that outperforms traditional methods in accuracy and automation, tailored for large cosmological surveys.

Findings

Achieves less than 20% error in separating galaxy types up to magnitude 22.

Outperforms classical C/A classification by more than a factor of two.

Comparable to space-based data in classification accuracy.

Abstract

We present a new non-parametric method to quantify morphologies of galaxies based on a particular family of learning machines called support vector machines. The method, that can be seen as a generalization of the classical CAS classification but with an unlimited number of dimensions and non-linear boundaries between decision regions, is fully automated and thus particularly well adapted to large cosmological surveys. The source code is available for download at http://www.lesia.obspm.fr/~huertas/galsvm.html To test the method, we use a seeing limited near-infrared ($K_s$ band, $2,16\mu m$) sample observed with WIRCam at CFHT at a median redshift of $z\sim0.8$. The machine is trained with a simulated sample built from a local visually classified sample from the SDSS chosen in the high-redshift sample's rest-frame (i band, $0.77\mu m$) and artificially redshifted to match the observing conditions. We use a 12-dimensional volume, including 5 morphological parameters and other caracteristics of galaxies such as luminosity and redshift. We show that a qualitative separation in two main morphological types (late type and early type) can be obtained with an error lower than 20% up to the completeness limit of the sample ($KAB\sim 22$) which is more than 2 times better that what would be obtained with a classical C/A classification on the same sample and indeed comparable to space data. The method is optimized to solve a specific problem, offering an objective and automated estimate of errors that enables a straightforward comparison with other surveys.