Open msqiqi10 opened 3 years ago
If I understand you correctly, you generate DeepWalk features separately for each graph. Our matching procedure is based on finding correspondences based on locally similar neighborhoods (which is not given in case of DeepWalk features as inputs). Instead of using those features, you can use the torch_geometric.transforms.OneHotDegree
transform. This should help the neural network to match nodes based on their structural similarities.
Hi:
Thank you for your great work!
Now I'm trying to run your code on a unfeatured, undirected graph with a feature vector generated from Deepwalk. But the performance is really poor. I'm wondering if you can please provide some ideas on how should I set up a feature vector? Or how could I improve the performace?
Thanks!