OmegaNeuron: Applying GravitySpy Similarity Methods to the Search for LIGO Glitch Witnesses

TL;DR

OmegaNeuron is a machine-learning tool that integrates GravitySpy's image similarity methods with Omega Scan's transient analysis to automate the identification of auxiliary channels that witness glitches, enabling faster analysis that improves glitch witness identification, enhancing both detector sensitivity and the reliability of gravitational-wave observations.

Abstract

Gravitational-wave (GW) astronomy has advanced our understanding of compact mergers through instruments like the Laser Interferometer Gravitational-Wave Observatory (LIGO). However, the extreme sensitivity required for these detections makes the instruments susceptible to short-duration transient noise, or glitches, which obscure GW data. Current tools such as Omega Scan and GravitySpy assist in identifying and classifying such noise, but are limited by manual inspection or dependence on large training sets. To address these challenges, we present \textit{OmegaNeuron}, a machine-learning tool that integrates GravitySpy's image similarity methods with Omega Scan's transient analysis to automate the identification of auxiliary channels that witness glitches. Applied to multiple glitch examples, OmegaNeuron consistently highlighted plausible witness channels and showed strong agreement with existing correlation tools, while providing clearer ranking through a quantitative similarity metric. Integrated into the \texttt{gwdetchar} package, OmegaNeuron enables faster analysis that improves glitch witness identification, enhancing both detector sensitivity and the reliability of gravitational-wave observations.

Figures (4)

• Figure 1: t-SNE visualization of the 200-dimensional feature space at GPS time 1403298875.817s. Points are colored by subsystem. Axes are in t-SNE units and not metrically meaningful. Clustering is observed within subsystems.
• Figure 2: OmegaNeuron applied to data from GW150914 (no known glitch). The t-SNE projection shows clustering of auxiliary channels by subsystem; axes are arbitrary. The strain spectrogram is marked with a yellow $\bigstar$, and the top four auxiliary spectrograms plotted in the bottom panels are indicated by $\blacktriangle$. The highest-similarity channels were from the CAL, SUS, and OMC subsystems, all of which are known unsafe couplings. Similarity values above 0.998 demonstrate internal consistency of the metric. A full list of acronyms is provided in Appendix \ref{['app:acronyms']}, Table I, and detailed rankings are provided in Appendix \ref{['app:ranks']} Table II.
• Figure 3: OmegaNeuron applied to a scattered-light glitch. The strain spectrogram is marked with a yellow $\bigstar$, and the four spectrograms shown below are indicated by $\blacktriangle$ in the t-SNE plot. Surrounding clusters in t-SNE plot correspond to the LSC and ASC subsystems, consistent with optical scattering pathways. The two highest-similarity channels closely resemble the strain spectrogram in frequency, duration, and energy, with similarity values above 0.925. Rankings for these and other channels are provided in Appendix \ref{['app:ranks']} Table III.
• Figure 4: OmegaNeuron applied to an unclassified glitch. The strain spectrogram is marked with a yellow $\bigstar$, and the top four auxiliary spectrograms plotted below are indicated by $\blacktriangle$ in the t-SNE plot. Nearest clusters in the t-SNE projection, consistent with the strain similarity rankings, include PSL, LSC, PEM, and OMC auxiliary channels. The first, second, and fourth highest-similarity spectrograms closely resemble the strain spectrogram in frequency, duration, and energy, with similarity values of 0.9914, 0.9905, and 0.9862, respectively. Rankings for these and other channels are provided in Appendix \ref{['app:ranks']} Table IV.