图电导、同步与一种新的瓶颈测度
Graph conductance, synchronization, and a new local bottleneck measure for decentralized network optimization
AI总结:
该研究针对图电导无法保证全节点同步流形稳定的问题,提出同步瓶颈测度$\Xi$量化任意节点子集与系统其余部分的流动瓶颈,其可控制一大类动力系统同步所需耦合强度,相关启发式方法可指导提升全局同步能力的去中心化策略。
摘要:
作为图上流动主要瓶颈的量化指标,电导这一网络属性在网络耦合动力系统的同步过程中发挥着重要作用。扩散耦合项不仅用于在网络化系统的节点间交换信息,最终还会耗散集体动态状态的熵,使其向同步流形稳定时可与单个动态节点关联的熵值降低。虽然图电导能够表征此类系统中维持绝大多数节点广泛同步所需的耦合强度,但它无法保证包含系统所有节点的同步流形的稳定性。我们定义了一种称为图的同步瓶颈的新测度,记为$\Xi$;这一新的网络属性量化了任意节点子集(无论其阶数大小)与网络化系统其余部分之间流动的极限瓶颈。该量确实控制了一大类动力系统中同步流形稳定所需的耦合强度。与电导的情况类似,求解该量具有组合复杂性,但基于此优化问题的启发式方法可为提升全局同步能力的去中心化策略提供指导。
英文摘要:
The two most prominent bottleneck measures in graph theory, commonly known as the isoperimetric number and conductance, are both referred to in the literature as Cheeger constants. While these measures are useful for assessing barriers to flow in networked systems, neither is sufficient to characterize the stability of complete synchronization, i.e. the dynamic convergence of every node in the system toward a synchronization manifold. Conductance, in particular, does provide an effective bound on the coupling strength required to achieve bulk synchronization, but it often fails to differentiate between many chimera states. The Fiedler vector, which is the eigenvector associated with the algebraic connectivity of the graph, is similarly often able to correctly identify the limiting cut, however, it can fail as well being a global relaxation, especially in the presence of small-set bottlenecks. Furthermore, the computations of these properties are NP-Hard and require global information about the network structure, which is often unrealistic for many complex systems. We define a normalized \textit{synchronization bottleneck ratio} for each bi-partitioning graph cut. The minimum of this ratio over all such cuts gives the \textit{synchronization bottleneck measure} as a global network property, and the associated cut better identifies the true limiting bottleneck for complete synchrony for a specific class of network coupled dynamics. Obtaining this global minimizer is also NP-hard, but due to the use of only local information in the argument, heuristics based on the ratio itself can guide decentralized strategies for improving the synchronizability in man-made networked systems; having relevance in power grid stability, distributed database coherence, and other applications where the alignment of every node is required for the proper functioning of the system.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。