BiGCN: A Bi-directional Low-Pass Filtering Graph Neural Network
Paper • 2101.05519 • Published
ANLP A3 Part 2
Custom PyTorch models trained on the supplied UPFD GossipCop splits. Contains 45 best-validation-macro-F1 checkpoints, their configurations, and a frozen experiment manifest. BiGCN follows Chen et al., arXiv:2101.05519, as linked in the assignment PDF. Use the source repository and bash scripts/part2/test_gnn.sh to regenerate the comparison tables.
Training source SHA256: ab04b42c9ef3ec0751d770466310079265f053306de7e25ab2276d06abdd6f8e Frozen manifest SHA256: 1cdb483fad854bdf3eb77f61aeb82e60fc2f49560f0dad5d1b10d4672f4b5a6a
These models are educational experiment artifacts; the dataset and cited papers retain their own licences.
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support