Deep View-Reduction TSK Fuzzy System: A Case Study on Epileptic EEG Signals Detection

In many practical applications, the fuzzy systems have been used due to the promising approximation accuracy and the high interpretability. Here, we proposed a novel multiview Takagi-Sugeno-Kang (TSK) fuzzy system in which a deep structure associating with a view-reduction mechanism are involved. The deep structure of each view is constructed by many basic components, i.e., the classic one-order TSK fuzzy systems which are linked in a layer by layer way using the stacked generalization principle. The view-reduction mechanism contains two parts: 1) A user-free parameter which is fixed according to the feature distribution is introduced to guild the view weight learning; 2) Views with noisy weights are automatically filtered by a reduction principle which is generated according to the training data. The proposed multi-view fuzzy system is finally applied for epileptic EEG signals detection.