TRUST-FS: Tensorized Reliable Unsupervised Multi-View Feature Selection for Incomplete Data

On November 12, 2025, the research community was introduced to TRUST-FS, a groundbreaking method for multi-view unsupervised feature selection (MUFS) aimed at overcoming persistent challenges in handling incomplete data. Traditional methods often focus solely on missing views, neglecting the broader issue of missing variables, which can lead to inaccurate results. TRUST-FS innovatively combines feature selection, missing-variable imputation, and view weight learning through an adaptive-weighted CP decomposition. This integrated approach not only enhances the accuracy of feature selection but also addresses the shortcomings of previous methodologies that treated imputation and selection as separate processes. Comprehensive experimental results have demonstrated the effectiveness of TRUST-FS, marking a significant step forward in the field of machine learning and data analysis, where accurate feature selection is crucial for the performance of various applications.