UPCommons. Portal del coneixement obert de la UPC
Publication
Contribución de congreso (ConferenceOutput)
NetXplain: Real-time explainability of graph neural networks applied to computer networks
Recent advancements in Deep Learning (DL) have revolutionized the way we can efficiently tackle complex optimization problems. However, existing DL-based solutions are often considered as black boxes due to their high inner complexity. As a result, there is still certain skepticism among the computer network industry about their practical viability to operate data networks. In this context, explainability techniques have recently emerged to unveil why DL models make each decision. This paper focuses on Graph Neural Network (GNN) models applied to computer networks, which have already shown outstanding performance in different network optimization tasks. We thus present NetXplain, a novel real-time explainability solution that uses a GNN to interpret the output produced by another GNN. In the evaluation, we apply the proposed explainability method to RouteNet –a GNN model that predicts end-to-end performance metrics in computer networks. We show that NetXplain operates more than 3 orders of magnitude faster than state-of-the-art explainability solutions when applied to networks up to 24 nodes, which makes this solution compatible with real-time applications. Moreover, it demonstrated strong generalization capabilities over different network scenarios unseen during training., This work has received funding from the European Union’s Horizon 2020 research and innovation programme within the framework of the NGI-POINTER Project funded under grant agreement No. 871528. This paper reflects only the author’s view; the EC is not responsible for any use that may be made of the information it contains. This work was also supported by the Spanish MINECO under contract TEC2017-90034-C2-1-R (ALLIANCE) and the Catalan Institution for Research and Advanced Studies (ICREA)., Peer Reviewed