PhD Seminar: Data-based strategies for the detection of continuous gravitational waves

Title: Data-based strategies for the detection of continuous gravitational waves
Author: Rodrigo Tenorio
Abstract:
Continuous gravitational waves (CW) are a (to date) undetected kind of gravitational wave signal characterized by its long-duration and extreme weakness when compared to the more familiar short-transient signals produced by the coalescence of compact objects.
This poses a problem to standard analysis techniques, for the required computing cost to deploy a matched-filtering (“nearly-optimal”) search greatly exceeds currently affordable computing allocations. To this end, more efficient (albeit non-optimal) approaches are customarily used, even though their concrete statistical behaviour is sometimes not properly understood.
In this work, we introduced two potential solutions to these problems: first, we launched a Kaggle data-analysis competition in order to propel the development of an end-to-end machine-learning search for CWs. This competition, currently undergoing, is sponsored by Google with a total of $25.000 in prizes, and its results will help us understand the suitability of machine-learning in this specific avenue of gravitational-wave data analysis. Second, we introduce a new approach to estimate the background distribution of general gravitational-wave searches, `distromax`. Using this technique, based on extreme value theory, we are able to establish statistically sound thresholds on detection statistics the specific distribution of which is poorly understood (if at all), potentially producing more sensitive searches at an unchanged computing cost. The outcome of these projects will provide decisive techniques the application of which could well help in the first detection of CW signals during the forthcoming runs of the Advanced LIGO detectors.