search

zhumeiqiBUPT / AM-GCN

AM-GCN: Adaptive Multi-channel Graph Convolutional Networks
MIT License
219 stars 46 forks source link

About the feature graph construction #7

Closed ZJJTSL closed 3 years ago

ZJJTSL commented 3 years ago

i wonder to know how to construct the feature graph, i can't make it clear how to get the 'citeseer.feature' file in the processdata.py , look forward for your help,thanks!

zhumeiqiBUPT commented 3 years ago

Hello, you can run process_data() in dataprocess.py to generate the 'citeseer.feature' file, using the original data files in gcn as input. But actually, we have already provided the feature file in citeseer.zip, and you just need to unzip it first. Then generate_knn() is used for constructing feature graph.

ZJJTSL commented 3 years ago

Hello, you can run process_data() in dataprocess.py to generate the 'citeseer.feature' file, using the original data files in gcn as input. But actually, we have already provided the feature file in citeseer.zip, and you just need to unzip it first. Then generate_knn() is used for constructing feature graph.

yeah,i've noticed the citeseer.zip,but i want to how to make it for a new dataset . In the dataprocess.py line 20, i don't know which file is opened, thanks!

Noval11 commented 3 years ago

Hello, you can run process_data() in dataprocess.py to generate the 'citeseer.feature' file, using the original data files in gcn as input. But actually, we have already provided the feature file in citeseer.zip, and you just need to unzip it first. Then generate_knn() is used for constructing feature graph.

yeah,i've noticed the citeseer.zip,but i want to how to make it for a new dataset . In the dataprocess.py line 20, i don't know which file is opened, thanks!

I have the same quesion. Looking forward to your answer! Thanks!

zhumeiqiBUPT commented 3 years ago

Hello, I think I've already given the explanation. The original data files in https://github.com/tkipf/gcn/tree/master/gcn/data is opened as input. You just need download and store them in data/cache/ folder before running dataprocess.py

Noval11 commented 3 years ago

Yes, you are right, I find it. Thank you!