Multilayer cellular learning automata: A computational model to solve multilayer infrastructure problems with its application in community detection for multilayer networks

Abstract

Multilayer networks are one type of complex systems that have multiple types of interactions among their nodes. Through the development and extension of the real-world networks, especially social and biological networks, understanding the structure and organization of multilayer systems become more significant. In this paper, we propose a new computational model based on learning automata to overcome the problems related to multilayer networks. In our proposed model, the whole network is modeled as a set of layers in which an irregular cellular learning automata (ICLA) is assigned to each node in a layer. The functionality of cellular learning automata (CLA) is dependent on two important features: local and global environments. In this model, the information from the other layers is evaluated as the global environments, and update rules are performed based on the information from the other layers. Finally, by the convergence of CLA in each layer the global environments cause the multilayer cellular learning automata (MLCLA) to cover all possible states in all nodes of layers. We apply this model to community detection in multilayer networks as an example of the application of the proposed model. To solve this, each CLA by the depth-first search algorithm detects the optimal community in a layer and then sends the information about the identified community to the next layer as a feature vector. The simulation results on multiple well-known multilayer datasets demonstrate the effectiveness and superiority of the proposed algorithm in terms of similarity and modularity measures, and the analysis results of the proposed algorithm in terms of Precision, Recall, and F1-score are higher than the other available community detection methods.