Congratulations to Ph.D. Student Kehua Yuan’s Paper “Robust Semi-supervised Feature Selection with Multi-granularity Zentropy Modeling” Accepted by TPAMI
Recently, the paper “Robust Semi-supervised Feature Selection with Multi-granularity Zentropy Modeling” by Ph.D. student Kehua Yuan, under the guidance of Prof. Duoqian Miao, has been accepted by IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), one of the top journals in artificial intelligence. The study addresses feature selection in high-dimensional weakly supervised data, proposing a model-agnostic multi-granularity zentropy framework (Ze-MGM) that captures hierarchical interactions among labels, decisions, and individual classes. Experiments show that Ze-MGM achieves superior classification performance and robustness compared with existing methods.
近日，实验室苗夺谦教授指导的2022级博士研究生苑克花的文章“Robust Semi-supervised Feature Selection with Multi-granularity Zentropy Modeling”被国际顶级期刊《IEEE Transactions on Pattern Analysis and Machine Intelligence》（TPAMI）正式录用。TPAMI是人工智能与模式识别领域公认的顶级期刊，长期位列中科院一区Top期刊，同时为CCF A类与CAAI A类推荐期刊，在国际学术界具有重要影响力。
该研究围绕高维弱监督数据场景下特征选择所面临的标注稀缺、不确定性与鲁棒性等关键问题，构建了一种多粒度不确定性建模框架（Ze-MGM）。区别于以往基于特定粗糙集或模糊集假设的粒计算方法，所提框架结合数据自身特性构建多粒度熵结构，系统刻画标签、决策与个案类别之间的层次交互关系（如图1所示），从而实现对高维弱监督数据不确定性的统一表征与推理。理论分析与实验结果表明，所提方法在分类性能和鲁棒性方面具有显著优势。
图1. 多粒度知识空间构建图