Estimating photometric redshifts with artificial neural networks

A new approach to estimating photometric redshifts - using Artificial Neural Networks (ANNs) - is investigated. Unlike the standard template-fitting photometric redshift technique, a large spectroscopically-identified training set is required but, where one is available, ANNs produce photometric redshift accuracies at least as good as and often better than the template-fitting method. The Bayesian priors on the underlying redshift distribution are automatically taken into account. Furthermore, inputs other than galaxy colours - such as morphology, angular size and surface brightness - may be easily incorporated, and their utility assessed. Different ANN architectures are tested on a semi-analytic model galaxy catalogue and the results are compared with the template-fitting method. Finally the method is tested on a sample of ~ 20000 galaxies from the Sloan Digital Sky Survey. The r.m.s. redshift error in the range z < 0.35 is ~ 0.021.