jknet

Representation Learning on Graphs with Jumping Knowledge Networks (JK-Net)

• Paper link: https://arxiv.org/abs/1806.03536
• Author's code repo: https://github.com/ShinKyuY/Representation_Learning_on_Graphs_with_Jumping_Knowledge_Networks. Note that the original code is implemented with Tensorflow for the paper.

Structure

How to run

Run with following (available dataset: "cora", "citeseer", "pubmed")

python jknet_trainer.py --dataset cora 

For details settings, please refer to here.

Results

TL_BACKEND="paddle" python jknet_trainer.py --dataset cora --mode max --lr 0.01 --n_epoch 170 --hidden_dim 32
TL_BACKEND="paddle" python jknet_trainer.py --dataset citeseer --mode max --lr 0.01 --n_epoch 200 --hidden_dim 64
TL_BACKEND="paddle" python jknet_trainer.py --dataset pubmed --mode cat --lr 0.01 --n_epoch 300 --hidden_dim 64 --itera_K 4
TL_BACKEND="tensorflow" python jknet_trainer.py --dataset cora --mode cat --lr 0.005 --n_epoch 200 --hidden_dim 64
TL_BACKEND="tensorflow" python jknet_trainer.py --dataset citeseer --mode cat --lr 0.01 --n_epoch 170 --hidden_dim 32
TL_BACKEND="tensorflow" python jknet_trainer.py --dataset pubmed --mode max --lr 0.01 --n_epoch 170 --hidden_dim 32
TL_BACKEND="torch" python jknet_trainer.py --dataset cora --mode max --lr 0.01 --n_epoch 200 --hidden_dim 16 
TL_BACKEND="torch" python jknet_trainer.py --dataset citeseer --mode cat --lr 0.01 --n_epoch 200 --hidden_dim 16 
TL_BACKEND="torch" python jknet_trainer.py --dataset pubmed --mode max --lr 0.1 --n_epoch 200 --hidden_dim 16

Dataset	Paper	Our(pd)	Our(tf)	Our(th)
cora	0.896(±0.005)	0.847(±0.01)	0.8584(±0.007)	0.872(±0.007)
citeseer	0.783(±0.008)	0.7554(±0.001)	0.761(±0.01)	0.769(±0.014)
pubmed		0.7782(±0.003)	0.7826(±0.005)	0.792(±0.005)