bits-pilani-hyd / bipartite-graph-learning

bipartite-graph-learning

Adversarial Representation Learning on Large-Scale Bipartite Graphs

Reproducibility

Preparation

pip3 install -r requirements.txt

Peproduciable Scripts Overview

	ABCGraph (Adversarial)	ABCGraph (MLP)	Node2Vec	GCN	GraphSAGE	GAE
Platform	MacOS/Linux	MacOS/Linux	Only Linux (*)	MacOS/Linux	MacOS/Linux	MacOS/Linux
Tencent	sh run_abcgraph_adv.sh tencent	sh run_abcgraph_mlp.sh tencent	sh run_node2vec.sh tencent	sh run_gcn.sh tencent	sh run_graphsage.sh tencent	N/A (*)
Cora	sh run_abcgraph_adv.sh cora	sh run_abcgraph_mlp.sh cora	sh run_node2vec.sh cora	sh run_gcn.sh cora	sh run_graphsage.sh cora	sh run_gae.sh cora
Citeseer	sh run_abcgraph_adv.sh citeseer	sh run_abcgraph_mlp.sh citeseer	sh run_node2vec.sh citeseer	sh run_gcn.sh citeseer	sh run_graphsage.sh citeseer	sh run_gae.sh citeseer
PubMed	sh run_abcgraph_adv.sh pubmed	sh run_abcgraph_mlp.sh pubmed	sh run_node2vec.sh pubmed	sh run_gcn.sh pubmed	sh run_graphsage.sh pubmed	sh run_gae.sh pubmed

Only Linux (*): For the Node2Vec model, its binary file is only ELF 64-bit LSB executable, x86-64, for GNU/Linux.

N/A (*): For the GAE model, the code of the original GAE paper can not simply applied to the large-scale bipartite graph due to the memory constrain. To apply GAE to the large-scale graph data is another research topic, so we don't report the result in the "Tencent" dataset. From the other datasets, we can see that our model's performance is better than the GAE model.

Background running:

# ABCGraph (Adversarial)
nohup sh run_abcgraph_adv.sh tencent > log_run_abcgraph_adv_tencent.txt 2>&1 &
nohup sh run_abcgraph_adv.sh cora > log_run_abcgraph_adv_cora.txt 2>&1 &
nohup sh run_abcgraph_adv.sh citeseer > log_run_abcgraph_adv_citeseer.txt 2>&1 &
nohup sh run_abcgraph_adv.sh pubmed > log_run_abcgraph_adv_pubmed.txt 2>&1 &

# ABCGraph (MLP)
nohup sh run_abcgraph_mlp.sh tencent > log_run_abcgraph_mlp_tencent.txt 2>&1 &
nohup sh run_abcgraph_mlp.sh cora > log_run_abcgraph_mlp_cora.txt 2>&1 &
nohup sh run_abcgraph_mlp.sh citeseer > log_run_abcgraph_mlp_citeseer.txt 2>&1 &
nohup sh run_abcgraph_mlp.sh pubmed > log_run_abcgraph_mlp_pubmed.txt 2>&1 &

# Node2Vec
nohup sh run_node2vec.sh tencent > log_run_node2vec_tencent.txt 2>&1 &
nohup sh run_node2vec.sh cora > log_run_node2vec_cora.txt 2>&1 &
nohup sh run_node2vec.sh citeseer > log_run_node2vec_citeseer.txt 2>&1 &
nohup sh run_node2vec.sh pubmed > log_run_node2vec_pubmed.txt 2>&1 &


# GCN
nohup sh run_gcn.sh tencent > log_run_gcn_tencent.txt 2>&1 &
nohup sh run_gcn.sh cora > log_run_gcn_cora.txt 2>&1 &
nohup sh run_gcn.sh citeseer > log_run_gcn_citeseer.txt 2>&1 &
nohup sh run_gcn.sh pubmed > log_run_gcn_pubmed.txt 2>&1 &

# GraphSAGE
nohup sh run_graphsage.sh tencent > log_run_graphsage_tencent.txt 2>&1 &
nohup sh run_graphsage.sh cora > log_run_graphsage_cora.txt 2>&1 &
nohup sh run_graphsage.sh citeseer > log_run_graphsage_citeseer.txt 2>&1 &
nohup sh run_graphsage.sh pubmed > log_run_graphsage_pubmed.txt 2>&1 &

# GAE
nohup sh run_gae.sh tencent > log_run_gae_tencent.txt 2>&1 &
nohup sh run_gae.sh cora > log_run_gae_cora.txt 2>&1 &
nohup sh run_gae.sh citeseer > log_run_gae_citeseer.txt 2>&1 &
nohup sh run_gae.sh pubmed > log_run_gae_pubmed.txt 2>&1 &

# AS-GCN
cd ASGCN/data
python tencent_dataset_loader.py
nohup python run_pubmed.py --dataset tencent > running_asgcn.txt