Glitch classification by ML

Yi Yang^1*, Chia-Jui Chou¹, Weichen Qin¹, Jiakai Zhang¹, Yufan Xie¹, Yu-Suan Lin²

¹School of Physical Science and Technology, ShanghaiTech University, Shanghai, China
²Department of Electrophysics, National Yang Ming Chiao Tung University, Hsinchu, Taiwan

* Presenter:Yi Yang, email:yangyi3@shanghaitech.edu.cn

I will present our recent results on glitch classification in gravitation wave detection for the KAGRA O4a run using a machine learning approach. After preprocessing the data, we generated spectrograms for a large set of glitches. We first developed an unsupervised model to classify the glitches, and then used these clustering results to label the data automatically, creating a training set for a supervised model. This semi-supervised strategy significantly improves the accuracy of glitch classification compared to using an unsupervised model alone, while avoiding the need for redundant manual labeling.

Keywords: gravitational wave, glitch classification, ML