Unified Residual Anomaly Detection for GW Time Series

The whole dataset used in the paper ‘Noise Reconstruction Driven Unified Anomaly Detection for Heterogeneous Transients in Gravitational Wave Time Series’ is provided here. Paper abstract: Detecting transient anomalies in gravitational-wave data is challenging due to complex noise. Existing template-based methods lack generalization. This paper proposes a noise-reconstruction-driven framework that amplifies deviations in residual space. Two reconstruction models and a multi-branch detector (Random Forest + InceptionTime) are developed. Experiments on LIGO data demonstrate robust unified detection across SNR conditions.