Fast Data-Centric Optimization of Nonlinear Dynamic Flows on Network System Suited for Big-Data and Extreme Computing

Abstract

—The network optimization problem, which is an optimization problem for a discrete system as a graph structure consisting of nodes and links, has been applied to several fields. Network optimization problems may be more difficult depending on the network scale and problem setting. In contrast, Big-Data & Extreme Computing (BDEC) is a method to process a large amount of data, such as observation data acquired by technologies such as 5G and IoT, by pouring them into abundant computing resources using high-speed cloud computing. In recent years, BDEC environments have become more common. Based on these considerations, this study aims to develop a network optimization method that combines AI and simulations to obtain a fast solution that is effective in BDEC and also to conduct basic studies on the effectiveness of the method by applying it to actual network optimization problems. Therefore, the method was applied to the inverse estimation problem, confirming its effectiveness. For the OD traffic estimation, heuristic solutions at high speed were obtained. Future developments comprise the inclusion of a reliable forward analysis to enhance the accuracy of the system reproduction, improvements in the sampling method and use of hyperparameter tuning method.