Institute of Astronomy

Fast follow-up using accelerated likelihood evaluation

SpeakerTalk DateTalk Series
Jonathan Gair (IoA)8 May 2013Institute of Astronomy Seminars


The calculation of posterior parameter distributions using Bayesian techniques requires repeated likelihood evaluations and can therefore be both computationally expensive and time-consuming. We will describe an approach to faster likelihood evaluation based on reduced order quadratures. A reduced spanning basis for the model space is first constructed and then a quadrature rule for that reduced basis is developed. The quadrature rule is typically cheaper to evaluate than the full likelihood, allowing accelerated follow-up. We discuss the application of these techniques to the follow-up of gravitational wave candidates identified by the LIGO interferometers.