Of Genes and Machines: application of a combination of machine learning tools to astronomy datasets

TL;DR

This paper demonstrates a machine learning approach combining Genetic Algorithms and Support Vector Machines to improve star/galaxy classification and photometric redshift estimation in astronomical datasets, achieving high accuracy and efficiency.

Contribution

It introduces a novel combined GA and SVM method for feature selection and parameter optimization in astronomy data analysis, outperforming existing classifiers.

Findings

Achieves 98% accuracy in star/galaxy classification down to i_P1=24.5

Yields photometric redshifts with sigma=0.013 in (1+z) for COSMOS galaxies

Outperforms the SExtractor spread_model classifier at faint magnitudes

Abstract

We apply a combination of a Genetic Algorithms (GA) and Support Vector Machines (SVM) machine learning algorithm to solve two important problems faced by the astronomical community: star/galaxy separation, and photometric redshift estimation of galaxies in survey catalogs. We use the GA to select the relevant features in the first step, followed by optimization of SVM parameters in the second step to obtain an optimal set of parameters to classify or regress, in process of which we avoid over-fitting. We apply our method to star/galaxy separation in Pan-STARRS1 data. We show that our method correctly classifies 98% of objects down to i_P1= 24.5, with a completeness (or true positive rate) of 99% for galaxies, and 88% for stars. By combining colors with morphology, our star/classification method yields better results than the new SExtractor classifier spread_model in particular at the faint end (i_P1>22). We also use our method to derive photometric redshifts for galaxies in the COSMOS bright multi-wavelength dataset down to an error in (1+z) of sigma=0.013, which compares well with estimates from SED fitting on the same data (sigma=0.007) while making a significantly smaller number of assumptions.