A view-reduction based multi-view TSK fuzzy system and its application for textile color classification

Abstract

Takagi–Sugeno–Kang (TSK) fuzzy systems are well known for their good balance between approximation accuracy and interpretability. Nowadays, multi-view TSK fuzzy systems have drawn world-wide attention since the multi-view data has emerged in many application scenarios. However, many multi-view data contains weak views which may bring negative influences on the task of pattern recognition. How to reduce the weak views and hence improve the performance of multi-view TSK fuzzy systems have become a hot topic. In this paper, we propose a view-reduction based multi-view TSK fuzzy system termed as VR-MV-TSK-FS in which two powerful mechanisms are developed: (1) during the collaborative learning of data in each view, an object-distribution-dependent parameter is defined to control the learning of the weight of each view. The parameter is not fixed by users, it is set according to the feature space in advance such that the learnt weight of each view indeed reflects the amount of pattern information involved in each view; (2) during the iteration of VR-MV-TSK-FS, weak views are automatically reduced by comparing the learnt weight with a fixed threshold which is also automatically set according to the number of objects and the dimension of the feature space. With the two mechanisms, a training algorithm is developed. Extensive experiments on synthetic datasets, UCI datasets and a case study of textile color classification indicate that the proposed algorithm can effectively reduce weak views and achieves better performance than other benchmarking algorithms.