面向多视图无监督特征选择的结构感知混合阶相似学习
机器学习
2025-12-01 v1
摘要
多视图无监督特征选择 (MUFS) 最近日益成为处理无标签多视图数据有效降维方法。然而,大多数现有方法主要使用一阶相似图来保留局部结构,常常忽略了二阶相似所能捕获的全局结构。此外,少数 MUFS 方法依赖预定义的二阶相似图,容易受到噪声和离群点的影响,导致特征选择性能不佳。本文提出了一种新颖的 MUFS 方法,称为结构感知混合阶相似学习用于多视图无监督特征选择 (SHINE-FS),以解决上述问题。SHINE-FS 首先学习一致性锚点及其对应的锚点图,以捕获锚点与样本之间的跨视图关系。基于获取的跨视图一致信息,它生成样本的低维表示,这有助于通过识别判别性特征来重建多视图数据。随后,它利用锚点-样本关系学习二阶相似图。 Furthermore, by jointly learning first-order and second-order similarity graphs, SHINE-FS constructs a hybrid-order similarity graph that captures both local and global structures, thereby revealing the intrinsic data structure to enhance feature selection. Comprehensive experimental results on real multi-view datasets show that SHINE-FS outperforms the state-of-the-art methods.
关键词
引用
@article{arxiv.2511.22656,
title = {Structure-aware Hybrid-order Similarity Learning for Multi-view Unsupervised Feature Selection},
author = {Lin Xu and Ke Li and Dongjie Wang and Fengmao Lv and Tianrui Li and Yanyong Huang},
journal= {arXiv preprint arXiv:2511.22656},
year = {2025}
}