Multi-scale locality preserving projection for partial multi-view incomplete multi-label learning.

Area of Science:

• Machine Learning
• Computer Vision
• Data Science

Background:

• Multi-view multi-label classification is crucial but challenged by incomplete data.
• Existing methods often neglect the importance of manifold structures in partial data.

Purpose of the Study:

• To propose MSLPP, a novel model for partial multi-view incomplete multi-label learning.
• To address the challenge of missing views and labels while preserving data's inherent structure.

Main Methods:

• MSLPP captures and integrates distance and similarity information from original and extracted feature spaces.
• Graph embedding preserves intrinsic structure and multi-scale information.
• A shielding strategy mitigates negative impacts of missing views.

Main Results:

• MSLPP effectively retains data's inherent structure during feature extraction.
• The shielding strategy accurately captures inherent data structure despite missing views.
• Experimental results demonstrate superior performance over existing methods on five datasets.

Conclusions:

• MSLPP offers a robust solution for partial multi-view incomplete multi-label classification.
• The proposed shielding strategy enhances model accuracy and data structure preservation.