Efficient Classification of Gravitational Wave Glitches with a Hybrid Ghost-Cnn Architecture

The Laser Interferometer Gravitational-Wave Observatory (LIGO) has been instrumental in detecting gravitational waves, which provide key insights into astrophysical phenomena. However, due to hardware limitations, the observatory is also susceptible to glitches and non-transient noise, which can mimic actual gravitational waves and hinder detection accuracy. This study addresses the challenge of classifying these glitches by proposing a hybrid Ghost Convolutional Neural Network (Ghost-Cnn) model, designed to distinguish glitches from true gravitational wave signals more efficiently than traditional models. We implemented and compared the performance of the hybrid Ghost-Cnn model against a baseline Convolutional Neural Network (CNN) on a dataset of glitch spectrograms by LIGO. While the base CNN achieved a respectable accuracy of 92%, the Ghost-Cnn model demonstrated a marginal improvement with a classification accuracy of 94%. Beyond accuracy, the hybrid model also outperformed the CNN across key metrics, including F1 score, precision, and recall, reflecting its superior ability to handle class imbalances and reduce false positives and negatives. A particularly notable outcome of this research is the 50% reduction in inference time achieved by the Ghost-Cnn compared to the base CNN. This substantial improvement in computational efficiency is critical for real-time applications, where fast and accurate glitch classification can enable continuous and more precise gravitational wave observation at LIGO. Overall, the Ghost-Cnn hybrid model boosts accuracy and efficiency, making it well-suited for real-time gravitational wave monitoring. Its rapid and precise glitch classification enhances detection reliability, contributing to astrophysical research and aiding the study of gravitational waves.