Predicting the Robustness of Real-World Complex Networks

Many real-world natural and social systems can be modeled as complex networks. As random failures and malicious attacks can seriously destroy the structure of complex networks, it is critical to ensure their robustness and maintain the functions. Generally, connectivity and controllability robustness are adopted to evaluate the performance of networked systems against external attacks and/or failures. A sequence of values is measured to dynamically indicate the network robustness with iterative node- or edge-removal. Calculating the robustness of large-scale real-world networks is usually time consuming, whereas deep-learning provides an efficient methodology to estimate network robustness performance. In this paper, a multi-convolutional neural network (CNN) method called Real-RP is designed to predict the robustness of real-world complex networks. Unknown real-world networks are first classified into known network categories, and their robustness performance is then predicted based on the knowledge of the specific network category trained using a substantial number of synthetic networks. Experimental results show that: 1) real-world complex networks can be classified by a CNN with high precision, and 2) the robustness performance of real-world networks can be predicted with lower average errors compared to existing methods.

Automated summary

This paper introduces the Real-RP, a multi-convolutional neural network method, for predicting the robustness of real-world complex networks. Experimental results show high precision in network classification and lower prediction errors compared to existing methods.