Abstract: In industrial anomaly detection, the traditional auto-associative kernel regression (AAKR) method suffers from low computational efficiency and interference from distant samples. To address ...