Multi-View Graph Learning by Joint Modeling of Consistency and Inconsistency.

Area of Science:

• Machine Learning
• Data Science
• Computer Vision

Background:

• Graph learning is effective for multi-view clustering by creating unified graphs.
• Existing methods often overlook inter-view inconsistencies, impacting performance on noisy data.

Purpose of the Study:

• To propose a novel multi-view graph learning framework.
• To explicitly model and address both multi-view consistency and inconsistency simultaneously.

Main Methods:

• Developed a unified objective function to learn consistent and inconsistent graph components.
• Designed an efficient optimization algorithm with linear time complexity.
• Framework adaptable to both similarity and dissimilarity graphs.

Main Results:

• The proposed framework learns a unified graph by fusing consistent view information.
• Demonstrated robustness and efficiency across 12 multi-view datasets.
• Achieved an approximate solution to an NP-hard optimization problem.

Conclusions:

• The novel framework effectively handles multi-view consistency and inconsistency.
• Offers a robust and efficient solution for multi-view clustering.
• Applicable to diverse graph types, enhancing its versatility.