Morphology Classification and Photometric Redshift Measurement of Galaxies

TL;DR

This paper uses SDSS data to classify galaxy morphology with k-means and improves photometric redshift estimation using neural networks, achieving high accuracy for different galaxy types.

Contribution

It introduces an unsupervised morphology classification combined with neural network-based redshift estimation, enhancing accuracy for galaxy samples.

Findings

Neural networks perform better with more input parameters.

Mixed accuracy for redshift estimation surpasses overall sample accuracy.

Achieved rms deviation of 0.0192 for optimal photometric redshift prediction.

Abstract

Based on the Sloan Digital Sky Survey Data Release 5 Galaxy Sample, we explore photometric morphology classification and redshift estimation of galaxies using photometric data and known spectroscopic redshifts. An unsupervised method, k-means algorithm, is used to separate the whole galaxy sample into early- and late-type galaxies. Then we investigate the photometric redshift measurement with different input patterns by means of artificial neural networks (ANNs) for the total sample and the two subsamples. The experimental result indicates that ANNs show better performance when the more parameters are applied in the training set, and the mixed accuracy $\sigma_{mix}=\sqrt{{\sigma_{early}}^2+{\sigma_{late}}^2}$ of photometric redshift estimation for the two subsets is superior to $\sigma_{z}$ for the overall sample alone. For the optimal result, the rms deviation of photometric redshifts for the mixed sample amounts to 0.0192, that for the overall sample is 0.0196, meanwhile, that for early- and late-type galaxies adds up to 0.0164 and 0.0217, respectively.