Photometric redshift estimation: An active learning approach

A long-lasting problem in astronomy is the accurate estimation of galaxy distances based solely on the information contained in photometric filters. Due to observational selection effects, the spectroscopic (source) sample lacks coverage throughout the feature space (e.g. colors and magnitudes) compared to the photometric (target) sample; this results in a clear mismatch in terms of photometric measurement distributions. We propose a solution to this problem based on active learning, a machine learning technique where a sampling strategy enables us to select the most informative instances to build a predictive model; specifically, we use active learning following a Query by Committee approach. We show that by making wisely selected queries in the target domain, we are able to increase our predictive performance significantly. We also show how a relatively small number of queries (spectroscopic follow-up measurements) suffices to improve the performance of photometric redshift estimators significantly.