Boundary heat diffusion classifier for a semi-supervised learning in a multilayer network embedding.

Area of Science:

• Network Science
• Machine Learning
• Data Mining

Background:

• High-quality annotations are scarce in many real-world applications.
• Learning techniques combining unlabeled and labeled data are increasingly important.
• Label propagation in multilayer networks is a challenging problem.

Purpose of the Study:

• To develop an efficient and effective label propagation algorithm for multilayer networks.
• To adapt the heat diffusion model for network label propagation.
• To address the scarcity of labeled data in network analysis.

Main Methods:

• Proposed a novel boundary-based heat diffusion algorithm.
• The algorithm is inspired by heat diffusion models used in machine learning.
• Guarantees a closed-form solution for efficient implementation.

Main Results:

• Experimental validation on synthetic and five real-world multilayer network datasets.
• The boundary-based heat diffusion algorithm demonstrated superior performance.
• Outperformed state-of-the-art methods in label propagation tasks.

Conclusions:

• The proposed boundary-based heat diffusion algorithm is effective for label propagation in multilayer networks.
• The method offers an efficient solution for leveraging unlabeled data.
• This approach shows significant benefits over existing state-of-the-art techniques.