To solve the problem of low generalization performance of traditional anomaly detection method AUC,a deep sup-port vector data description anomaly detection method based on hybrid kernel maximum correlation entropy is proposed.Firstly,the mean square error loss function is replaced by the maximum correlation entropy loss function.Then,the original fixed kernel of the maximum correlation entropy loss function is improved by hybrid kernel to improve the robustness and generalization performance of the model.Finally,the anomaly detection method based on MM-DSVDD is constructed,and the high AUC value is achieved in both MNIST and Fashion MNIST datasets,indicating that the MM-DSVDD anomaly detection method has high accuracy and appli-cation prospect.
mixed kenelsmaximum correlation entropyanomaly detectioncondition monitoringdeep support vector da-ta description
万方数据
维普
