GNN-SimBench

The training_dataset and testing_dataset are Graph Neural Network datasets generated with synthetic data from OMNET++, following the specifications of the winner of the https://bnn.upc.edu/challenge/gnnet2022/ challenge. The training split lets researchers develop and fine-tune GNN architectures across diverse topologies, routing strategies and traffic loads, providing adjacency lists, node/link features, traffic matrices, simulator settings, run logs and ground-truth labels for exact case replays defined in cases.json. The testing split mirrors this format but with labels withheld, enabling objective benchmarking of model generalisation on unseen scenarios while still allowing deterministic re-creation of every simulation case. This information can also be found in the next link, please check it for updates: PREDICT 6G / GNN Dataset GitLab (uc3m.es) This work has been funded by the European Commission Horizon Europe SNS JU PREDICT-6G (GA 101095890) Project.