Self-supervised reconstruction of transients in data from space-borne gravitational-wave detectors

概要

Space-based gravitational-wave (GW) data may contain transient signals whose waveform morphologies are not known in advance. Extracting these signals is important for characterizing new sources and mitigating instrumental anomalies. However, existing deep neural network (DNN)-based extraction approaches rely on clean training targets and waveform-class-specific examples, which may limit their applicability when the transient morphology is not specified in advance. This work develops a Noise2Nois…