Network Optimization Based on Analysis for Real and Virtual Spaces

Virtual network topologies changes due to the spatio-temporal demand change by users and the node and/or link failures. Since such topology change affects the resource allocation on the physical network, we need to efficiently allocate the resources in terms of the network characteristics and connectivity in both the physical and virtual networks. To address this issue, this research aims at realizing the optimal network control based on both the graph neural network and reinforcement learning by combining analysis and prediction results of real space with network characteristics in the virtual space.