Discrimination reveals reconstructability of multiplex networks from partial observations

Area of Science:

• Network science
• Complex systems analysis
• Information theory

Background:

• Link prediction in monoplex networks is established, but multiplex network reconstruction lacks robust mathematical tools.
• Existing methods struggle to reconstruct multiplex network structures from aggregate topology and partial link data.
• Accurate reconstruction of multiplex networks is crucial for understanding complex systems.

Purpose of the Study:

• To develop a theoretical and computational framework for reconstructing multiplex networks.
• To identify key factors influencing the accuracy of multiplex network reconstruction.
• To propose optimal strategies for allocating observed links across layers to enhance reconstruction.

Main Methods:

• Developed a probability space with maximum likelihood estimation for structural inference.
• Introduced 'discrimination' as an entropy-based indicator of inter-layer differences.
• Utilized empirical data from biological, social, and engineered systems, alongside synthetic networks.

Main Results:

• The proposed framework enables accurate reconstruction of multiplex networks.
• Network discrimination quantifies the reconstructability of multiplex networks.
• Optimal link allocation strategies were designed based on discrimination values.
• Theoretical findings were validated across diverse real-world and synthetic network datasets.

Conclusions:

• The developed framework provides a novel mathematical tool for multiplex network reconstruction.
• Discrimination is a critical determinant of multiplex network reconstructability.
• The findings facilitate improved strategies for network data acquisition and analysis in multiplex systems.