基于卷积神经网络的复杂网络鲁棒性优化

Robust Optimization of Complex Networks Based on Convolutional Neural Networks

近年来,复杂网络的鲁棒性优化问题引起了人们广泛关注.目前,基于单一测度的鲁棒性优化已经取得很大成就,但复杂网络暴露在外会同时受到多种破坏,从而导致网络系统崩溃无法正常运行.因此构建能同时抵抗节点和连边攻击下的网络结构尤为重要,但在优化过程中复杂网络多目标能控性鲁棒性计算非常耗时.为了解决该问题,文中提出了一种基于卷积神经网络(CNN)的多目标复杂网络能控性鲁棒性优化进化算法.该方法在3种合成网络和两种真实网络上的实验结果表明,在优化过程中使用CNN来评估网络的能控性鲁棒性不仅能降低优化时间,也能辅助进化算法找到抗击能力更强的网络结构.

In recent years,the robustness optimization problem of complex networks has attracted widespread attention.,At present,robust optimization based on a single measure has achieved great success,but complex networks exposed to external factors can suffer from multiple damages,leading to network system crashes and inability to operate normally.Therefore,it is particularly important to build a network structure that can resist both node and edge attacks simultaneously,but in the optimization process,calculating the multi-objective controllability and robustness of complex networks is very time-consuming.To address this issue,a multi-objective complex network controllability and robustness optimization evolutionary algorithm based on Convolutional Neural Network(CNN)is proposed in this paper.The experimental results of this method on three synthetic networks and two real networks show that using CNN to evaluate the controllability and robustness of the network during the optimization process can not only reduce optimization time,but also assist evolutionary algorithms in finding network structures with stronger resistance.