An Evolutionary Multitasking Method for Multi-Objective Critical Node Detection on Interdependent Networks

Abstract

The critical node detection problem (CNDP) on interdependent networks has attracted the attention of researchers recently. However, as the number of network layers increases leading to the expansion of the search space, it prevents the existing work from obtaining high-quality solutions with limited computational resources. In fact, there is connection information between each single-layer network and the interdependence network. In this regard, this paper proposes an evolutionary multitasking method in which knowledge transfer between single-layer and multi-layer network tasks is fully utilized to better solve the CNDP on interdependent networks. Specifically, the CNDP on interdependent networks is firstly transformed into a bi-objective problem (BICND). Then, the critical node detection (CND) task for multi-layer network and the CND task for single-layer networks are established in the proposed method to solve the BICND problem. Meanwhile, this paper proposes a new knowledge transfer mechanism based on the node coupling relationship of interdependent networks. This knowledge transfer mechanism not only ensures the transfer of useful information between multiple tasks, but also helps them to obtain higher quality solutions. The experimental results show that the proposed method can effectively search for high-quality solutions and has better performance than other single-objective and multi-objective methods.