First-Order Sparse TSK Nonstationary Fuzzy Neural Network Based on the Mean Shift Algorithm and the Group Lasso Regularization
Bingjie Zhang,
Jian Wang,
Xiaoling Gong,
Zhanglei Shi,
Chao Zhang,
Kai Zhang,
El-Sayed M. El-Alfy,
Sergey V. Ablameyko
Affiliations
Bingjie Zhang
School of Mathematical Sciences, Dalian University of Technology, Dalian 116024, China
Jian Wang
College of Science, China University of Petroleum (East China), Qingdao 266580, China
Xiaoling Gong
College of Control Science and Engineering, China University of Petroleum (East China), Qingdao 266580, China
Zhanglei Shi
College of Science, China University of Petroleum (East China), Qingdao 266580, China
Chao Zhang
School of Mathematical Sciences, Dalian University of Technology, Dalian 116024, China
Kai Zhang
School of Petroleum Engineering, China University of Petroleum (East China), Qingdao 266580, China
El-Sayed M. El-Alfy
Fellow SDAIA-KFUPM Joint Research Center for Artificial Intelligence, Interdisciplinary Research Center of Intelligent Secure Systems, Information and Computer Science Department, King Fahd University of Petroleum and Minerals, Dhahran 31261, Saudi Arabia
Sergey V. Ablameyko
Faculty of Applied Mathematics and Computer Science, Belarusian State University, 220030 Minsk, Belarus
Nonstationary fuzzy inference systems (NFIS) are able to tackle uncertainties and avoid the difficulty of type-reduction operation. Combining NFIS and neural network, a first-order sparse TSK nonstationary fuzzy neural network (SNFNN-1) is proposed in this paper to improve the interpretability/translatability of neural networks and the self-learning ability of fuzzy rules/sets. The whole architecture of SNFNN-1 can be considered as an integrated model of multiple sub-networks with a variation in center, variation in width or variation in noise. Thus, it is able to model both “intraexpert” and “interexpert” variability. There are two techniques adopted in this network: the Mean Shift-based fuzzy partition and the Group Lasso-based rule selection, which can adaptively generate a suitable number of clusters and select important fuzzy rules, respectively. Quantitative experiments on six UCI datasets demonstrate the effectiveness and robustness of the proposed model.
