Usage of a Graph Neural Network for Large-Scale Network Performance Evaluation
Cen Wang, Noboru Yoshikane, Takehiro Tsuritani
Abstract
To quickly and accurately perform network evaluation, we propose a graph convolutional network-based performance evaluation method for ultralarge-scale networks. The learning results show that our method outperforms the fully connected network and convolutional neural network in the prediction error of the end-to-end latency and network throughput. In addition, we show that our method is significantly less time-consuming than traditional methods.
Topics & Concepts
Computer scienceConvolutional neural networkLatency (audio)GraphArtificial intelligenceNetwork performanceArtificial neural networkMachine learningTheoretical computer scienceComputer networkTelecommunicationsAdvanced Graph Neural NetworksComplex Network Analysis TechniquesSoftware-Defined Networks and 5G