A general approach to domain adaptation with applications in astronomy

Abstract

The ability to build a model on a source task and subsequently adapt this model to a new target task is a pervasive need in many astronomical applications. The problem is generally known in the machine learning field as transfer learning, where domain adaptation is a popular scenario. An example is to build a predictive model on spectroscopic data to identify Type Ia supernovae (SNe Ia), while subsequently trying to adapt such a model to photometric data. In this paper we propose a new general approach to domain adaptation which does not rely on the proximity of source and target distributions. Instead we simply assume a strong similarity in model complexity across domains, and use active learning to mitigate the dependence on source examples. Our work leads to a new formulation for the likelihood as a function of empirical error using a theoretical learning bound; the result is a novel mapping from generalization error to a likelihood estimation. Results using two real astronomical problems, SN Ia classification and identification of Mars landforms, show two main advantages of our approach: increased performance accuracy and substantial savings in computational cost.