Feature-Embedded Evolutionary Algorithm for Network Optimization

Network optimization problems are usually NP-hard and evolutionary algorithms (EAs) are good at these problems. However, direct implementation of EA, ignoring the network properties or problem features, is common which usually results in a low-efficient algorithm. In this paper, a framework of feature-embedded evolutionary algorithm is outlined and the idea is demonstrated by embedding the nodal centralities in a genetic algorithm (GA) for solving the node selection problem in pinning control. Simulation results confirm that the new design outperforms existing deterministic schemes in terms of solution quality, and significant improvements are also noticed as compared to classical GA. The results also shed light on high efficient feature-embedded EAs in solving other challenging optimization problems.