Challenging the generalization capabilities of Graph Neural Networks for network modeling - BNN-UPC/Papers GitHub Wiki

Demo paper

J. Suárez-Varela, S. Carol-Bosch, K. Rusek, P. Almasan, M. Arias, P. Barlet-Ros, A. Cabellos-Aparicio.

Proceedings of the ACM SIGCOMM 2019 Conference Posters and Demos, pp. 114-115, Beijin, China, August 2019.

Link to paper: Here(https://doi.org/10.1145/3342280.3342327)

Citation

You can cite this paper as follows:

@inproceedings{gnn-sigcomm,
  title={Challenging the generalization capabilities of Graph Neural Networks for network modeling},
  author={Su\'arez-Varela, Jos\'e and Carol-Bosch, Sergi and Rusek, Krzysztof and Almasan, Paul and Arias, Marta and Barlet-Ros, Pere and Cabellos-Aparicio, Albert},
  booktitle={Proceedings of the ACM SIGCOMM 2019 Conference Posters and Demos},
  pages={114-115},
  year={2019}
}

Abstract

Today, network operators still lack functional network models able to make accurate predictions of end-to-end Key Performance Indicators (e.g., delay or jitter) at limited cost. Recently, a novel Graph Neural Network (GNN) model called RouteNet was proposed as a cost-effective alternative to estimate the per-source/destination pair mean delay and jitter in networks. Thanks to its GNN architecture that operates over graph-structured data, RouteNet revealed an unprecedented ability to learn and model the complex relationships among topology, routing and input traffic in networks. As a result, it was able to make performance predictions with similar accuracy than resource-hungry packet-level simulators even in network scenarios unseen during training. In this demo, we will challenge the generalization capabilities of RouteNet with more complex scenarios, including larger topologies with a wider variety of routing configurations and traffic intensities than in the original work's evaluation.

The source code, the training/evaluation datasets and the delay model already trained used in this paper are available at the following link:

[Code]