智能系统学报2025，Vol.20Issue(3)：557-570,14.DOI:10.11992/tis.202408015
基于伪标签细化的域适应TSK模糊分类器
Domain adaptive Takagi-Sugeno-Kang fuzzy classifier based on pseudo-label refinement
摘要
Abstract
The Takagi-Sugeno-Kang(TSK)fuzzy classifier(FC)has been widely applied to various fields owing to its excellent classification performance and interpretability.To address the degradation of the generalization performance of this TSK TSK FC caused by the differences in the distributions of the training and test samples,a domain adaptive(DA)pseudo-label refinement(PLR)-based TSK FC(DA-TSK-PLR-FC)is proposed.This classifier leverages the nonlinear and linear mapping capabilities of the antecedent and consequent parts in fuzzy rules to construct a fuzzy shared feature space for source and target domain data.In this fuzzy shared feature space,graph-based random walking and label filter-ing refinement were applied to enhance the pseudo-label quality in the target domain,thereby enhancing the effective-ness of the domain alignment.Further,extensive experiments using multiple public datasets reveal that the proposed DA-TSK-PLR-FC achieves reliable classification performance and good interpretability.关键词
域适应/Takagi-Sugeno-Kang模糊分类器/随机游走/伪标签细化/模糊共享特征空间/无监督学习/模糊规则/迁移学习Key words
domain adaptation/Takagi-Sugeno-Kang fuzzy classifier/random walking/pseudo-label refinement/fuzzy shared feature space/unsupervised learning/fuzzy rule/transfer learning分类
信息技术与安全科学引用本文复制引用
张馨匀,周琳家,程煜婷,邱成羽,谢宇航,陈秀,张远鹏..基于伪标签细化的域适应TSK模糊分类器[J].智能系统学报,2025,20(3):557-570,14.基金项目
江苏省研究生研究实践创新计划项目(KYCX24_3561) （KYCX24_3561）
中国博士后科学基金资助项目(2023T160342) （2023T160342）
国家自然科学基金项目(82072019). （82072019）
