Deep learning ensemble for real-time gravitational wave detection of spinning binary black hole mergers

• DOI: 10.1016/j.physletb.2020.136029
• Publication Date: 2020-12-17T01:24:55Z
• Authors: Wei Wei; Asad Khan; E.A. Huerta; Xiaobo Huang; Minyang Tian
• Keywords: Black hole (networking)

We introduce the use of deep learning ensembles for real-time, gravitational wave detection spinning binary black hole mergers. This analysis consists training independent neural networks that simultaneously process strain data from multiple detectors. The output these is then combined and processed to identify significant noise triggers. have applied this methodology in O2 O3 finding clearly mergers open source available at Gravitational-Wave Open Science Center. also benchmarked performance new by processing 200 hours source, advanced LIGO August 2017. Our findings indicate our approach identifies real sources with a false positive rate 1 misclassification every 2.7 days searched data. A follow up misclassifications identified them as glitches. ensemble represents first class network classifiers are trained millions modeled waveforms describe quasi-circular, spinning, non-precessing, Once fully trained, processes faster than real-time using 4 NVIDIA V100 GPUs.