AEDNet: Adaptive Edge-Deleting Network For Subgraph Matching

This repository is the official implementation of 'AEDNet: Adaptive Edge-Deleting Network For Subgraph Matching'.

Official Download: Here can download official paper.

Free Download: Here can download paper.

Official Download: Here can download official paper.

Free Download: Here can download paper.

Requirements

• python3.7
• pytorch==1.9.0
• dgl==0.8.0
• networkx==2.6.2
• numpy==1.21.5
• matplotlib==3.4.2

This code repository is heavily built on DGL, which is a DEEP GRAPH LIBRARY for Graph Computation. Please refer here for how to install and utilize the library.

Datasets

Generate Data

There are some samples in './data/'. You should generate data before training.

To generate the Synthetic Data, run this command:

python creatData.py

Processing Data

You can use dgraph.__getitem __() in dataSet.py to process one sample and then use collate() in dataSet.py to batch data.

See one data sample'interior structure

You can use this command to see one data sample's interior structure.

from dgl.data.utils import save_graphs, get_download_dir, load_graphs

graph_pair_path = './data/COX2/train/0.bin'  ## one data sample's path
graph_pair, label_dict = load_graphs(graph_pair_path)
graph_data = graph_pair[0]  ##  one sample's data graph in DGL form
graph_query = graph_pair[1]  ##  one sample's query graph in DGL form
label = label_dict['glabel']  ##  Ground-Truth matching relatinship
print(graph_data, graph_query, label)

Training

To train the model(s) in the paper, run this command:

python train.py

Reference

If you find our paper/code is useful, please consider citing our paper:

@article{lan2023aednet,
  title={AEDNet: Adaptive Edge-Deleting Network For Subgraph Matching},
  author={Lan, Zixun and Ma, Ye and Yu, Limin and Yuan, Linglong and Ma, Fei},
  journal={Pattern Recognition},
  volume={133},
  pages={109033},
  year={2023},
  publisher={Elsevier}
}