An Attention-Based Deep Kernel Learning Identification Method for Turntable Servo System(Code&Data)

To establish an accurate model of the turntable servo system under small-sample conditions, an attention-based deep kernel learning identification method is proposed to overcome the limited capability of conventional deep kernel learning in capturing key features and its weak interpretability in feature extraction. By enhancing the feature extraction process, both interpretability and feature representation are improved, thereby increasing the accuracy of nonlinear factors identification. To further improve overall modeling accuracy without relying on difficult-to-measure nonlinear data, the linear and nonlinear models of the system are obtained using an alternating compensation strategy for linear and nonlinear model errors.