Multitask Twin Pinball Kernel Matrix Classifier for Pattern Classification

Given the sensitivity of traditional matrix classifiers to noisy signals, a new classification method called multitask twin pinball kernel matrix classifier (MTTPKMC) is proposed. MTTPKMC first designs a feature mapping kernel (FMK) to protect the structural information of matrix samples and then introduces a pinball constraint term to reduce the sensitivity of the model to noise. In addition, the proposed algorithm is integrated with multitask learning (MTL) framework to fully utilize the correlation information between tasks, thereby further improving the classification performance of the model. To validate the effectiveness of the MTTPKMC method described above, experimental verification is conducted on two mechanical fault datasets, and the results show that the proposed method has superior diagnostic performance.