25 January 2019 in IEEE Transactions on Fuzzy Systems
The superior interpretability and uncertainty modeling ability of Takagi-Sugeno-Kang fuzzy system (TSK FS) make it possible to describe complex nonlinear systems intuitively and efficiently. However, classical TSK FS usually adopts the whole feature space of the data for model construction, which can result in lengthy rules for high-dimensional data and lead to degeneration in interpretability. Furthermore, for highly nonlinear modeling task, it is usually necessary to use a large number of rules which further weakens the clarity and interpretability of TSK FS. To address these issues, a concise zero-order TSK FS construction method, called ESSC-SL-CTSK-FS, is proposed in this paper by integrating the techniques of enhanced soft subspace clustering (ESSC) and sparse learning (SL). In this method, ESSC is used to generate the antecedents and various sparse subspace for different fuzzy rules, whereas SL is used to optimize the consequent parameters of the fuzzy rules, based on which the number of fuzzy rules can be effectively reduced. Finally, the proposed ESSC-SL-CTSK-FS method is used to construct con-cise zero-order TSK FS that can explain the scenes in high-dimensional data modeling more clearly and easily. Experiments are conducted on various real-world datasets to confirm the advantages.
9 March 2020 in IEEE Transactions on Fuzzy Systems
As the first attempt, this study considers how to provide a design methodology for style clustering on stylistic data where each cluster depends on both the similarities between data samples and its latently or apparently distinguishable style. By taking our previous fuzzy k plane clustering algorithm F-KPC as the basic framework, a fuzzy style k-plane clustering algorithm S-KPC is proposed to have its distinctive merits: (1) the nuances between styles of clusters can be well identified by using the proposed twofold data representation. That is to say, style matrices are used to express the structure hence style information of each cluster, while the augmentation of the original features of data with enhanced nodes is taken as an abstract representation so as to move the manifold structure of data apart. Such a twofold data representation can make us realize S-KPC readily in an incremental way; (2) By means of alternating optimization strategy, the objective function of S-KPC can be optimized such that each discriminant function of each cluster shares the advantages of both simple regression models and functional-link neural networks. Extensive experiments on synthetic and real-world datasets demonstrate that S-KPC has comparable clustering performance to several compared methods on the adopted ordinary datasets, and yet obviously outperforms them on stylistic datasets.
12 April 2021 in IEEE Transactions on Fuzzy Systems
While our recent deep fuzzy classifier DSA-FC, which stacks adversarial interpretable Takagi–Sugeno–Kang fuzzy subclassifiers, shares its promising classification, its training speed will become very slow and even intolerable for large-scale datasets, due to successive training on all training samples with their random gradient based updates along each layer of its stacked structure. In order to circumvent this bottleneck issue, a fast training algorithm FTA is developed in this study by downsizing fuzzy rules with the proposed gradient guided learning for each subclassifier at each layer of DSA-FC on large-scale datasets. The core of FTA is to assure fast training of each subclassifier at each layer of DSA-FC, which first generates first-order smooth gradient guided information by means of the proposed top- k fuzzy rules selected from all fuzzy rules in each subclassifier, and then quickly updates the current inputs in terms of such information, which will be taken as the inputs of the subclassifier at the next layer. Our theoretical analysis reveals that the proposed gradient guided learning indeed enhances the generalization capability of a deep fuzzy classifier with or without adversarial attacks on outputs. Experimental results on large datasets demonstrate that FTA indeed trains the deep fuzzy classifier DSA-FC quickly with enhanced generalization capability.
27 March 2019 in IEEE Transactions on Fuzzy Systems
Many practical scenarios have demanded that we should classify unlabeled data more accurately based on both physical features (e.g. colour, distance, or similarity) and implicit style features of data. As most extant classification algorithms classify unlabeled data based only on their physical features, they become weak in achieving expected classification results for many scenarios. To work around this drawback in this paper, a novel classification method (FuCM) from the perspective of fuzzy social network based on both physical and implicit style features of data is proposed. Based on the proposed fuzzy social network and its dynamics about fuzzy influences of nodes, FuCM comprises two stages. In its training stage, after the fuzzy social network has been built, it learns the topological structure, reflecting physical features and implicit style features, of data by carrying out fuzzy influence dynamics in the built network. In its prediction stage, both physical and implicit style features of data are effectively integrated to yield the double structure efficiency characterized by fuzzy influences of nodes. FuCM classifies unlabeled data according to the strongest connection measure, based on the proposed double structure efficiency. FuCM does not assume that both data distribution and the classification by physical features or by both physical and implicit style features of data must be known in advance. Thus, it is a novel unified classification framework in this sense. In contrast to all the nine comparative methods, FuCM experimentally demonstrates its comparable classification performance on most synthetic, UCI and KEEL datasets, which can be well classified based only on physical features of data. Furthermore, it displays distinctive superiority on five case studies, where satisfactory classification certainly depends on both physical and implicit style features.
19 December 2023 in IEEE Transactions on Emerging Topics in Computational Intelligence
Takagi–Sugeno–Kang (TSK) fuzzy inference has been widely used in approximating uncertain nonlinear systems because of its high interpretability and precision. However, TSK fuzzy inference models are usually constructed on data in a supervised manner for classification and regression problems. As the first attempt, a novel Takagi–Sugeno–Kang fuzzy clustering (TSK-FC) method from the perspective of unsupervised learning is presented in this paper. The distinctive characteristic of TSK-FC lies in that it groups data points into different clusters directly by fuzzy inference on the TSK fuzzy rules. Therefore, the comprehensibility of the clustering procedure and clustering results of TSK-FC can be obtained. Specifically, all input features in TSK-FC on the IF-parts of all fuzzy rules are randomly assigned with interpretable linguistic terms, which ensures the interpreta-bility of TSK-FC. In addition, TSK-FC explores the potential clusters in data patterns by iteratively updating K consequent parameter vectors of all fuzzy rules in an unsupervised manner, which can represent the K obtained clusters. Particularly, all data points from the same cluster will be as close as possible to the cluster center, and as far as possible from the other cluster centers in the iterative update procedure. Moreover, the optimization problem of TSK-FC can be transformed into a standard difference of convex functions using the concave–convex procedure (CCCP), which can be easily solved by introducing an iterative algorithm. Experimental results on synthetic and benchmark datasets as well as four case studies validate the effectiveness of the proposed fuzzy rule-based TSK-FC. The statistical results on all experiments indicate the feasibility of TSK-FC as a clustering tool. The code of TSK-FC is available at https://github.com/gusuhang10/TSK-FC .
23 March 2022 in Information Sciences