H2GCN-Pytorch

This repo is a pytorch implementation of H2GCN raised in the paper "Beyond Homophily in Graph Neural Networks: Current Limitations and Effective Designs" . Original tensorflow implementation can be found here.

Requirement

This project should be able to run without any modification after following packages installed.

pytorch
networkx
torch-sparse
torch-geometric

Tutorial

Run train.py

usage: train.py [-h] [--seed SEED] [--without-relu] [--epochs EPOCHS] [--lr LR] [--k K] 
                [--wd WD] [--hidden HIDDEN] [--dropout DROPOUT] 
                [--patience PATIENCE] [--dataset DATASET] [--gpu GPU] 
                [--split-id SPLIT_ID]
                                                                                                                                                                                                                   
optional arguments:                                                                                                                                                                                                
  -h, --help           show this help message and exit                                                                                                                                                             
  --seed SEED          seed                                                                                                                                                                                        
  --without-relu       disable relu for all H2GCN layer                                                                                                                                                                                           
  --epochs EPOCHS      number of epochs to train                                                                                                                                                                   
  --lr LR              learning rate                                                                                                                                                                               
  --k K                number of embedding rounds                                                                                                                                                                  
  --wd WD              weight decay value                                                                                                                                                                          
  --hidden HIDDEN      embedding output dim                                                                                                                                                                        
  --dropout DROPOUT    dropout rate
  --patience PATIENCE  patience for early stop
  --dataset DATASET    dateset name
  --gpu GPU            gpu id to use while training, set -1 to use cpu
  --split-id SPLIT_ID  the data split to use

Custom dataset

All dataset used in this repo were forked from repo geom-gcn. Custom dataset should fit following format :

PROJECT_ROOT/new_data/DATASET_NAME/
out1_graph_edges.txt            # format for each lines : SRC_NODE DST_NODE
out1_node_feature_label.txt     # format for each rows  : NODE_ID f0,f1,···

Use model.py

If you only want to use model.py separately, you need to pass two matrix to forward function while training.

adj : torch.sparse.Tensor.
x : torch.FloatTensor.