HongyangGao / Graph-U-Nets

Pytorch implementation of Graph U-Nets (ICML19)
http://proceedings.mlr.press/v97/gao19a/gao19a.pdf
GNU General Public License v3.0
513 stars 100 forks source link

How to do node classification? #2

Closed Yfhu1103 closed 5 years ago

Yfhu1103 commented 5 years ago

Hello, I only see your code for graph classfication, But not for node classification, also, your data folder does not contain dataset like citeceer or cora, I want to know what's going on? Thank you for your reply.

HongyangGao commented 5 years ago

Hi, the code for node classification is written in TensorFlow. I may update them in a branch of this code base. Also this code can be applied to node classification by removing CNN parts.

AppleXY commented 5 years ago

Hello,I have seen that you said I need remove CNN layer to make node classification ,can you tell me the file name that contains the CNN parts?Thank you !

Yfhu1103 commented 5 years ago

Hello,I have seen that you said I need remove CNN layer to make node classification ,can you tell me the file name that contains the CNN parts?Thank you !

Hi, I have read the code and I think the CNN part is in "network.py", and the CNN architecture is according to the paper in AAAI 2018 "An End-to-End Deep Learning Architecture for Graph Classification", If you want node classification only, codes in "ops.py" is enough.

psp3dcg commented 4 years ago

Hello @Mrslock , have you succeeded training GraphUNet on cora dataset(node classification) with ops.py? I have tried modified the code, but the result is not so good. Training loss decreased but validation loss is not(floats a bit, almost constant).:(

psp3dcg commented 4 years ago

@HongyangGao , Hello, could you please release the graph unet code for node classification? Thanks a lot:)

liuxy1103 commented 3 years ago

For node classification,how do I deal with the issue about batchsize?