Graph neural networking challenge 2021: Creating a scalable network digital twin
Graph Neural Networks (GNN) have produced groundbreaking applications in different fields where data is fundamentally structured as graphs (e.g., chemistry, recommender systems). In the field of computer networks, this new type of neural networks is being rapidly adopted for a variety of relevant networking use cases, particularly for those involving large and complex graphs (e.g., performance modeling, routing optimization, resource allocation in wireless networks).
This talk presents the “Graph Neural Networking challenge 2021”. The second edition of this competition brings a fundamental limitation of existing GNNs: their lack of generalization capability to larger graphs. In order to achieve production-ready GNN-based solutions, we need models that can be trained in network testbeds of limited size, and then be able to operate with guarantees in any real customer network, which are often 10x larger in number of nodes. In this challenge, participants are asked to train their GNN models in small network scenarios (up to 50 nodes), and then test their accuracy in networks of increasing size not seen before, up to 300 nodes. Solutions with better scalability properties will be the winners.