Cost-constrained network dismantling using quadratic evolutionary algorithm for interdependent networks

The dismantling and protection of networks is a significant problem that has wide-ranging applications and attracts many researchers. Most current studies only focus on single-layer or one-to-one interdependent networks. However, this paper considers the more realistic case where the links between layers in interdependent networks are one-to-many, and the networks’ robustness is studied accordingly. To solve the problem of dissolving interdependent networks under the premise of heterogeneous costs, we propose a cost-constrained elite quadratic evolutionary algorithm (CCEEA) based on cost constraints. Based on the network’s prior information, the initial optimal feasible solutions derived from four classical algorithms are regarded as the initial elite individuals of CCEEA. The set of attack nodes is then continuously updated interactively according to a new evolutionary mechanism with flexible updates so that the combination of nodes in the final set of attack nodes can maximally facilitate the disintegration of the network. We conducted experiments on a series of representative networks and showed that on synthetic networks, the CCEEA algorithm outperforms the other four state-of-the-art attack strategies by more than 13% in terms of disintegration, which is up to 25% higher. In particular, it can be up to more than 90% higher in real networks.