jiawozhong/mgmn

Code for "Multilevel Graph Matching Networks for Deep Graph Similarity Learning"

Multilevel Graph Matching Networks for Deep Graph Similarity Learning

1. Description

In this paper, we propose a Multilevel Graph Matching Network (MGMN) framework for computing the graph similarity between any pair of graph-structured objects in an end-to-end fashion. MGMN consists of a node-graph matching network for effectively learning cross-level interactions between each node of one graph and the other whole graph, and a siamese graph neural network to learn global-level interactions between two input graphs.

1.1 citation:

Xiang Ling, Lingfei Wu, Saizhuo Wang, Tengfei Ma, Fangli Xu, Chunming Wu and Shouling Ji, Multilevel Graph Matching Networks for Deep Graph Similarity Learning, IEEE Transactions on Neural Networks and Learning Systems (TNNLS), 2021

 @article{ling2021multilevel,
   title={Multilevel Graph Matching Networks for Deep Graph Similarity Learning},
   author={Ling, Xiang and Wu, Lingfei and Wang, Saizhuo and Ma, Tengfei and Xu, Fangli and Liu, Alex X and Wu, Chunming and Ji, Shouling},
   journal={IEEE Transactions on Neural Networks and Learning Systems (TNNLS)},
   publisher={IEEE},
   volume={},
   number={},
   pages={},
   url={https://ieeexplore.ieee.org/document/9516695},
   year={2021}
 }

1.2 glance:

├─── src
│    ├─── model
│    │    ├─── __init__.py
│    │    ├─── DenseGGNN.py
│    │    ├─── DenseGraphMatching.py
│    ├─── __init__.py
│    ├─── cfg_config.py
│    ├─── cfg_train.py
│    ├─── data.py
│    ├─── ged_config.py
│    ├─── ged_train.py
│    ├─── simgnn_utils.py
│    ├─── utils.py
├─── data
│    ├─── CFG
│    │    ├─── ...
│    ├─── GED
│    │    ├─── ...
├─── ...

2. How to run the codes

2.1 The Graph-Graph Classification Task

(1) Prepare the dataset for the classification task.

Datasets for the graph-graph classification task is provided in data/CFG directory.

(2) Specify some hyper-parameters for classification tasks in src/cfg_config.py

(3) Train and test the model by running the following command:

cd src
python cfg_train.py

2.2 The Graph-Graph Regression Task

(1) Prepare the dataset for regression tasks

Data for the graph-graph regression task are placed in /data/GED directory. All the files required by our codes can be downloaded following instructions in this repo. Please make sure you have downloaded all the 3 directories required by our code: data, save, and result.

After downloading these files, please put them under /data/GED, which is the default data folder by our configuration, or you can also specify your own data directory.

An example directory structure is:

data
├─── GED
│   ├── data
│   │    ├── AIDS700nef/
│   │    ├── LINUX/
│   ├── result
│   │    ├── aids700nef/
│   │    ├── linux/
│   ├── save
│   │    ├── dist_mat/
│   │    ├── aids700nef_ged_astar_gidpair_dist_map.pickle
│   │    ├── linux_ged_astar_gidpair_dist_map.pickle

(2) Specify some hyper-parameters for regression tasks in src/ged_config.py

(3) Train and test the model by running the following command:

cd src
python ged_train.py