Incomplete Multiview Nonnegative Representation Learning With Graph Completion and Adaptive Neighbors.

Area of Science:

• Machine Learning
• Data Mining
• Computer Vision

Background:

• Incomplete multiview clustering (IMC) is challenging due to missing data across views.
• Existing methods often overlook the importance of these missing views when modeling correlations.

Purpose of the Study:

• To propose a novel Incomplete Multiview Nonnegative representation learning model with Graph completion and Adaptive neighbors (IMNGA).
• To simultaneously address common graph learning, missing graph completion, and consensus nonnegative representation learning.

Main Methods:

• IMNGA utilizes common and incomplete graphs to reconstruct completed graphs for each view.
• It employs nonnegative representation learning by factorizing completed and incomplete graphs.
• Adaptive neighbors are incorporated to enhance graph completion and representation.

Main Results:

• The proposed IMNGA model effectively reconstructs completed graphs by considering both available and missing view information.
• It achieves a superior consensus representation that adheres to common graph constraints.
• Experimental results demonstrate IMNGA's effectiveness in incomplete multiview clustering tasks.

Conclusions:

• IMNGA offers a robust solution for incomplete multiview clustering by integrating graph completion and nonnegative representation learning.
• The method successfully models correlations across all views, including those with missing data.
• IMNGA outperforms existing state-of-the-art methods in incomplete multiview clustering.