Layer reconstruction and missing link prediction of a multilayer network with maximum a posteriori estimation

Area of Science:

• Network Science
• Graph Theory
• Computational Biology

Background:

• Multilayer networks consist of distinct layers representing different interaction types among shared nodes.
• Some layers within a multilayer network can exhibit structural similarities and interdependencies.
• Understanding and reconstructing these complex network structures is crucial for various scientific domains.

Purpose of the Study:

• To propose a novel maximum a posteriori (MAP)-based method for studying and reconstructing the structure of a target layer in a multilayer network.
• To develop a technique for identifying structurally similar layers within a multilayer network.
• To enable accurate prediction of missing links in network reconstruction.

Main Methods:

• Characterizing nodes within the target layer using vectors to compute edge weights.
• Employing eigenvector centrality for comparing network structures and detecting similar layers.
• Utilizing identified similar layers to obtain conjugate prior parameters for the MAP algorithm.

Main Results:

• The proposed MAP-based method successfully reconstructs the target layer of multilayer networks.
• The method demonstrates effectiveness in predicting missing links, even when a substantial number of links are absent.
• Validation on two real-world multilayer networks confirms the method's promising performance.

Conclusions:

• The developed MAP estimation technique offers a robust solution for multilayer network reconstruction.
• The approach is particularly valuable for analyzing networks with incomplete or missing link data.
• This work advances the field of network science by providing a powerful tool for structural analysis and prediction.