Unsupervised multi-view feature selection via attentive hierarchical bipartite graphs with optimizable graph filter

Xia Dong, Danyang Wu, Yiping Ke, Feiping Nie

Abstract: Highlights•Proposes parameter-free hierarchical bipartite graph construction for multi-view learning.•Introduces an attentive fusion strategy for multi-view, multi-hierarchical graph fusion.•Develops an optimizable graph filter to capture low-frequency components for feature selection.•Proposes an alternating optimization algorithm with comprehensive theoretical analysis.•Outperforms 15 state-of-the-art methods across 8 real-world datasets in feature selection.