BayesWave: a novel method for detecting un-modeled gravitational wave bursts

The principal challenge of gravitational wave (GW) data analysis is to separate true GW events from non-gaussian noise artifacts. The LIGO-Virgo Burst group has developed several algorithms for detecting un-modeled GW bursts that may be associated with supernova, gamma-ray bursts (GRB), or new physics. These existent algorithms fit for the signals, but do not include explicit models for the non-gaussian detector noise. We describe the BayesWave algorithm, which uses Bayesian model selection techniques to simultaneously fit coherent GW signals across a network of detectors along with non-gaussian and non-stationary noise features, or glitches, in each detector. BayesWave can identify instrument artifacts for detector characterization studies and produce `cleaned' data streams for use by template based searches, such as those for compact binary coalescence.