We introduce waveletnet, a wavelet-based neural-network architecture for identifying non-Gaussian noise in gravitational-wave data and reducing its impact on candidate ranking. Traditionally, convolutional neural networks (CNNs) have been widely used as a flexible machine learning method to mitigate non-Gaussian noise. However, training CNNs requires many data samples, especially when the input data segments are long. Glitches that mimic high-mass black hole signals are empirically known to have a waveletlike structure. We exploit this property in waveletnet by using simple neural networks to learn the best family of wavelets to model glitches in the LIGO-Virgo-KAGRA O3 data. Due to its simplicity, our framework is significantly more sample-efficient than CNNs. As a use case, we build upon the trigger inference using the extended strain representation (TIER) method from [arXiv:2507.08318] and show waveletnet can improve the performance of search pipelines. We take potential GW candidates from the pipeline, and then downweight the candidates having noisy strain regions in their vicinity [i.e., within $t_{\text{candidate}}\pm \mathcal{O}(10)\sec$]. We use our framework in a modular way: we provide an output score that can be added to the pipeline’s existing detection statistic for the candidates. We test our method using candidates from the ias-hm search pipeline and find peak improvements of $\sim 15\%$ in the search sensitive volume, concentrated at high total masses and asymmetric mass ratios.

Close

PrevNext

View figure in article

• General relativity
• Gravitational waves