单层与多层网络鲁棒性分析与优化

发布时间：2018-05-09 14:46

本文选题：网络鲁棒性 + 边攻击策略　；参考：《西安电子科技大学》2015年硕士论文

【摘要】：复杂网络充斥在人们生活的每个角落,时时刻刻影响着人们的生活,国内外学者从未停止对它研究的步伐。复杂网络的结构是交互信息和传输资源的载体,它的鲁棒性与否直接关系着国家的经济和人民的生活。所以,如何找到复杂网络鲁棒性较高的结构,是近年来国内外学者对复杂网络研究的重点之一。其中,在优化网络结构中,大部分的研究都是保证了网络中每个节点的度不改变,这样做是为了对现存的网络结构进行优化。而现实社会中的网络结构在不停的变化,网络中的节点的度随时可能发生着变化。同时,网络中节点之间的路径是非常关键的,它是任何信息传输的渠道。再者,社会中的网络都是相互关联的,构成了多层网络(Network of Network,NON),它们之间的连接性是网络相互依靠的重要组成部分。本论文首先对复杂网络在对不同边攻击下的鲁棒性进行了优化分析研究,其次对网络中的自然连接性进行了优化,再次对多层网络的连接进行了定量的分析。本文的主要工作如下:1:在优化网络结构的鲁棒性过程中,设定了一个新的约束准则,并且采取了三种不同类型的边攻击策略来对网络结构进行攻击。实验中采取了四种不同类型的人工模拟网络和一个实际网络作为初始网络,设计了一个简单的启发式的优化方法(Sample Heuristic,SH)来对这些网络在不同边攻击下进行鲁棒性的优化。实验结果表明对于不同结构的初始网络,在遭受不同的边攻击的情况下,经过SH方法优化后的结果都非常接近,但是他们的网络结构却不同。并且,我们发现在经过一种攻击策略优化后的网络结构,在经受其它攻击策略的攻击下,结构并不鲁棒,甚至会低于初始值。2:采取了自然连接性作为衡量网络鲁棒性的标准,设计了一种基于模拟退火算法的无标度网络自然连接性的优化方法,并且把保证优化前后网络的度不发生改变作为优化网络结构的前提条件。在模拟退火算法中,设计了一种巧妙的边交换策略(Increase Loop 3,IL-3)。实验中采取了两种实际存在的网络结构和人工模拟生成的无标度网络对设计的模拟退火算法的性能进行了测试;将模拟退火算法与现有的几种算法进行了比较,实验结果表明模拟退火算法能更好的得到鲁棒性较高的网络结构。3:考虑现实社会中复杂网络之间的相互连接,对四种不同类型的网络所构成的NON的鲁棒性进行了定量的分析,给出人们在以后创建NON网络中的一些建议。实验中使用人工模拟生成的四种不同类型的网络结构,对它们两两连接生成的NON网络中的节点分别进行任意攻击和恶意攻击。实验结果表明,在抵御恶意攻击的方面,规则网络和小世界网络之间所构成的NON有很好的鲁棒性;在抵御任意攻击的方面,无标度网络与其它网络之间所构成的NON有很好的鲁棒性。
[Abstract]:The complex network is filled in every corner of people's life and affects people's life all the time. Scholars at home and abroad have never stopped the pace of research on it. The structure of complex network is the carrier of interactive information and transmission resources, and its robustness is directly related to the national economy and people's life. Therefore, how to find the structure of complex networks with high robustness is one of the most important researches on complex networks at home and abroad in recent years. Among them, in the optimization of network structure, most of the research is to ensure that the degree of each node in the network does not change, in order to optimize the existing network structure. However, the network structure in real society is constantly changing, and the degree of nodes in the network may change at any time. At the same time, the path between nodes in the network is very critical, it is any channel of information transmission. Moreover, the networks in the society are interrelated, forming the network of network of network. The connectivity between them is an important part of the network dependence. In this paper, firstly, the robustness of complex networks under different edge attacks is studied, then the natural connectivity in the network is optimized, and then the connection of multi-layer networks is quantitatively analyzed. The main work of this paper is as follows: in the process of optimizing the robustness of network structure, a new constraint criterion is set up, and three different kinds of edge attack strategies are adopted to attack the network structure. In the experiment, four different types of artificial simulation networks and a real network are used as the initial networks, and a simple heuristic optimization method is designed to optimize the robustness of these networks under different edge attacks. The experimental results show that for the initial networks with different structures, the results of SH optimization are very close to each other under different edge attacks, but their network structure is different. Furthermore, we find that under the attack of other attack strategies, the network structure is not robust and even lower than the initial value. 2: natural connectivity is adopted as the criterion to measure the robustness of the network. An optimization method for the natural connectivity of scale-free networks based on simulated annealing algorithm is designed, and the premise of optimizing the network structure is to ensure that the degree of the network does not change before and after the optimization. In the simulated annealing algorithm, an ingenious edge switching strategy is designed. In the experiment, two kinds of practical network structure and scale-free network are used to test the performance of the simulated annealing algorithm, and the simulated annealing algorithm is compared with the existing algorithms. The experimental results show that the simulated annealing algorithm can get a more robust network structure. 3. Considering the interconnection of complex networks in real society, the robustness of NON composed of four different types of networks is analyzed quantitatively. Some suggestions for creating NON networks in the future are given. In the experiment, four different types of network structures are generated by artificial simulation, and the nodes in the NON network generated by their pairwise connections are attacked arbitrarily and maliciously respectively. The experimental results show that the NON formed between the rule network and the small-world network has good robustness in resisting malicious attacks, and in the aspect of resisting arbitrary attacks, The NON between scale-free network and other networks has good robustness.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：O157.5

【相似文献】