EAGER: SSDIM: Multiscale Methods for Generating Infrastructure Networks

这个早期概念探索性研究补助金(AGER)项目的最终目标是开发一个跨域多规模网络和相互依赖的关键基础设施(ICI)生成器，该生成器捕获真实网络的许多特征，并包含任意大小程度的随机性。从已知或假想网络的样本开始，生成器将合成网络和ICI的集合，这些集合平均将在其结构的多个尺度上保留一组不同的拓扑和物理设计属性。这些属性将包括中心性、可选性、路径长度、集群、流量、所需物理容量和模块化的几个衡量标准。这种方法在多个尺度上在这些属性中的许多属性中引入了跨整体的无偏见的可变性，这创造了合成系统所需的现实感。这些模型将包括在拓扑和物理设计中纳入的人为因素组件。将开发一个算法和启发式工具箱，用于生成基础设施和综合信息系统的综合网络，并将分发生成的基准。这项工作的结果将促进ICIS的模拟、策略测试和决策等任务，这些任务是通过网络生成算法的根本进步来实现的。工具箱将使用模块化方法进行设计，使其能够发展并应用于其他领域。由计算机科学、行为科学、土木工程、网络科学和公共卫生方面的专业知识组成的跨学科调查团队将理想地支持现实合成数据生成的目标。将利用网络分析、大数据系统、机器学习、水网络和组织科学的观点和方法来开发包括网络分析、优化和统计分析技术相结合的方法工具包。由此产生的产品将包括开发的算法和为广泛的科学界传播的生成的合成数据集。

The ultimate goal of this EArly-concept Grant for Exploratory Research (EAGER) project is to develop a crossdomain multiscale network and interdependent critical infrastructure (ICI) generator that captures many features of real networks and incorporates an arbitrarily large or small degree of stochasticity. Starting from samples of known or hypothesized networks, the generator will synthesize ensembles of networks and ICIs that will preserve, on average, a diverse set of topological and physical design properties at multiple scales of its structure. These properties will include several measures of centrality, assortativity, path lengths, clustering, flows, required physical capacities, and modularity. This approach introduces an unbiased variability across the ensemble in many of these properties at multiple scales which creates a desired realism of the synthesized system. The models will include human factor components incorporated in both topological and physical designs. A toolbox of algorithms and heuristics for generating synthetic networks of infrastructures and ICIs will be developed, and generated benchmarks will be disseminated. Outcomes of this work will facilitate such tasks as simulation, policy testing and decision making for ICIs enabled by fundamental advancement in network generation algorithms. The toolbox will be designed using a modular approach that will allow it to evolve and to be applied in other domains. The interdisciplinary team of investigators comprising expertise in computer science, behavioral science, civil engineering, and network science and public health will ideally support the goal of realistic synthetic data generation. Perspectives and methodological approaches from network analysis, big data systems, machine learning, water networks, and organizational sciences will be brought to bear to develop a toolkit of methods that include a combination of network analytics, optimization, and statistical analysis techniques. The resulting products will include developed algorithms and generated synthetic datasets disseminated for a broad scientific community.