Page 190 - 《软件学报》2020年第11期

软件学报 ISSN 1000-9825, CODEN RUXUEW E-mail: jos@iscas.ac.cn
Journal of Software,2020,31(11):3506−3518 [doi: 10.13328/j.cnki.jos.005846] http://www.jos.org.cn
©中国科学院软件研究所版权所有. Tel: +86-10-62562563

∗
基于训练空间重构的多模块 TSK 模糊系统

1,2
1
1
周 塔 , 邓赵红 , 蒋亦樟 , 王士同 1
1
(江南大学 数字媒体学院,江苏 无锡 214122)
2
(江苏科技大学 电气与信息工程学院,江苏 张家港 215600)
通讯作者: 周塔, E-mail: jkdzhout@just.edu.cn

摘 要: 利用重构训练样本空间的手段,提出一种多训练模块 Takagi-Sugeno-Kang(TSK)模糊分类器 H-TSK-FS.它
具有良好的分类性能和较高的可解释性,可以解决现有层次模糊分类器中间层输出和模糊规则难以解释的难题.为
了实现良好的分类性能,H-TSK-FS 由多个优化零阶 TSK 模糊分类器组成.这些零阶 TSK 模糊分类器内部采用一种
巧妙的训练方式.原始训练样本、上一层训练样本中的部分样本点以及所有已训练层中最逼近真实值的部分决策
信息均被投影到当前层训练模块中,并构成其输入空间.通过这种训练方式,前层的训练结果对后层的训练起到引导
和控制作用.这种随机选取样本点、在一定范围内随机选取训练特征的手段可以打开原始输入空间的流形结构,保
证较好或相当的分类性能.另外,该研究主要针对少量样本点且训练特征数不是很大的数据集.在设计每个训练模块
时采用极限学习机获取模糊规则后件参数.对于每个中间训练层,采用短规则表达知识.每条模糊规则则通过约束方
式确定不固定的输入特征以及高斯隶属函数,目的是保证所选输入特征具有高可解释性.真实数据集和应用案例实
验结果表明,H-TSK-FS 具有良好的分类性能和高可解释性.
关键词: TSK 模糊系统;多模块训练;解释能力;极限学习机
中图法分类号: TP181

中文引用格式: 周塔,邓赵红,蒋亦樟,王士同.基于训练空间重构的多模块 TSK 模糊系统.软件学报,2020,31(11):3506−3518.
http://www.jos.org.cn/1000-9825/5846.htm
英文引用格式: Zhou T, Deng ZH, Jiang YZ, Wang ST. Multi-module TSK fuzzy system based on training space reconstruction.
Ruan Jian Xue Bao/Journal of Software, 2020,31(11):3506−3518 (in Chinese). http://www.jos.org.cn/1000-9825/5846.htm
Multi-module TSK Fuzzy System Based on Training Space Reconstruction

1
1
1,2
ZHOU Ta , DENG Zhao-Hong , JIANG Yi-Zhang , WANG Shi-Tong 1
1
(School of Digital Media, Jiangnan University, Wuxi 214122, China)
2
(School of Electrical and Information Engineering, Jiangsu University of Science and Technology, Zhangjiagang 215600, China)
Abstract: A multi-training module Takagi-Sugeno-Kang (TSK) fuzzy classifier, H-TSK-FS, is proposed by means of reconstruction of
training sample space. H-TSK-FS has good classification performance and high interpretability, which can solve the problems of existing
hierarchical fuzzy classifiers such as the output and fuzzy rules of intermediate layer that are difficult to explain. In order to achieve
enhanced classification performance, H-TSK-FS is composed of several optimized zero-order TSK fuzzy classifiers. These zero-order
TSK fuzzy classifiers adopt an ingenious training method. The original training sample, part of the sample of the previous layer and part
of the decision information that most approximates the real value in all the training layers are projected into the training module of the
current layer and constitute its input space. In this way, the training results of the previous layers play a guiding and controlling role in the
training of the current layer. This method of randomly selecting sample points and training features within a certain range can open up the

∗ 基金项目: 国家自然科学基金(61772239, 61702225, 61572236); 江苏省自然科学基金(BK20181339)
Foundation item: National Natural Science Foundation of China (61772239, 61702225, 61572236); Natural Science Foundation of
Jiangsu Province (BK20181339)
收稿时间: 2018-04-12;修改时间: 2018-12-04;采用时间: 2019-03-28