Abstract

RouteNet is a novel network model based on Graph Neural Network (GNN) that can understand the complex relationship between topology, routing, and input traffic to produce accurate estimates of the per-source/destination per-packet delay distribution and loss. RouteNet leverages the ability of GNNs to learn and model graph-structured information and as a result, the model can generalize over arbitrary topologies, routing schemes and traffic intensity. In the evaluation,the RouteNet can predict accurately the delay distribution (mean delay and jitter) and loss even in topologies, routing and traffic unseen in the training.

Bio

Krzysztof Rusek defended his Ph.D. Thesis on queuing theory in 2016 at AGH University of Science and Technology, Kraków, Poland. Prior that He has worked as a system administrator and machine learning engineer in the research group focused on processing and protection of multimedia content. Currently, he is an assistant professor at AGH. He also works as a data scientist for Barcelona Neural Networking Center, IT advisor for the department of Atsronomy of Jagiellonian University and a Member of the Board of an AI-focused spin-of company. His main research interests are performance evaluation of telecommunications systems, machine learning, and data analysis. Currently, he is working on the applications of Graph Neural Networks and probabilistic modeling for performance evaluation of communications systems.