In this paper, we study convergence of coupled dynamical systems on convergent sequences of graphs to a continuum limit. We show that the solutions of the initial value problem for the dynamical system on a convergent graph sequence tend to that for the nonlocal diffusion equation on a unit interval, as the graph size tends to infinity. We improve our earlier results in [Arch. Ration. Mech. Anal., 21 (2014), pp. 781--803] and extend them to a larger class of graphs, which includes directed and undirected, sparse and dense, random and deterministic graphs.
There are three main ingredients of our approach. First, we employ a flexible framework for incorporating random graphs into the models of interacting dynamical systems, which fits seamlessly with the derivation of the continuum limit. Next, we prove the averaging principle for approximating a dynamical system on a random graph by its deterministic (averaged) counterpart. The proof covers systems on sparse graphs and yields almost sure convergence on time intervals of order $\log n,$ where $n$ is the number of vertices. Finally, a Galerkin scheme is developed to show convergence of the averaged model to the continuum limit.
The analysis of this paper covers the Kuramoto model of coupled phase oscillators on a variety of graphs including sparse Erdős-R{\' e}nyi, small-world, and power law graphs.
On solutions to the continuum version of the Kuramoto model with identical oscillators
仓本模型提供了一个具体的数学实现涌现同步人口的相位耦合振荡器。自从Kuramoto的出版物\textit{Oscillations, Waves, and Turbulence}以来,研究人员一直致力于更好地表征溶液动力学。在本文中,我们结合的方法的特征与迭代技术证明的连续版本的Kuramoto模型的特殊情况下,振子是相同的和描述的振子密度方面的限制投影特征的解的存在性和唯一性。该模型的特征,反过来,提供了全球渐近分析系统通过子和超级解决方案。我们相信,这里开发的独特的模型分析方法有可能在广泛研究的仓本模型的其他更复杂版本上产生新的结果。
The semilinear heat equation on sparse random graphs
使用$L^p$ -图子理论(Borgs et al,2014),我们推导并严格证明了稀疏随机图上微分方程系统的连续极限。具体地说,我们证明了离散模型的初值问题的解可以近似为一个适当的非局部扩散方程的解。我们的研究结果适用于一系列不同的物理,生物,社会和经济网络的空间扩展的动力学模型。重要的是,我们的假设涵盖了许多重要的现实网络中的网络拓扑结构。特别地,我们导出了幂律图上耦合动力系统的连续极限。后者是这项工作的主要动机。
How Low Can You Go? Active Learning for Sparse Model Discovery in the Ultra-Low-Data Limit
Department of Mechanical Engineering, University of Washington(华盛顿大学机械工程系);NSF AI Institute in Dynamic Systems, University of Washington(华盛顿大学NSF动态系统人工智能研究所);Department of Aeronautics, Imperial College London(伦敦帝国理工学院航空系)