Is GNN useful in your graph computing job?

[qingwen220]

GeaFlow has provided a graph inference framework currently, I'm wondering if it's helpful to provide some GNN algorithms. Could you share some ideas?

[Manveen777]

Yes, that could definitely be helpful. GeaFlow’s inference framework can already serve graph models, so adding common GNNs would make it more practical. Some useful candidates could be:
GCN for node classification
GraphSAGE for inductive learning on large/dynamic graphs
GAT for attention-based heterogeneous graphs
GIN for tasks that need stronger expressive power

[qingwen220]

Thanks for replying. If you are experienced in the part, we could co-work on it. We are planning to provide some algorithms in the next few months.