About this document
Hybrid-Label Feature Selection in MVML by sizheduan36 is a document available to read on EtoBox.
The document presents a method called Double-Layer Hybrid-Label Identification (DHLI) for multi-view multi-label feature selection, addressing the challenges of noise and specific characteristics in data from multiple sources. The proposed approach introduces a unified loss function that separates common, specific, and noisy labels, while also incorporating a novel regularization paradigm to enhance learning direction and optimize feature selection. Experimental results demonstrate the effectiveness of DHLI
- Author
- sizheduan36
- Language
- EN