Awesome

ImGAGN: Imbalanced Networks Embedding via Generative Adversarial Graph Networks

This is our Pytorch implementation for the paper:

Liang Qu, Huaisheng Zhu, Ruiqi Zheng, Yuhui Shi, and Hongzhi Yin. 2021. ImGAGN:Imbalanced Network Embedding via Generative Adversarial Graph Networks. In Proceedings of the 27th ACM SIGKDD Conference on Knowledge Discovery and Data Mining (KDD ’21), August 14–18, 2021, Virtual Event, Singapore. ACM, New York, NY, USA, 9 pages. https://doi.org/10.11453447548.3467334

Introduction

This work presents a generative adversarial graph network model, called ImGAGN to address the imbalanced classification problem on graphs. It introduces a novel generator for graph structure data, named GraphGenerator, which can simulate both the minority class nodes’ attribute distribution and network topological structure distribution by generating a set of synthetic minority nodes such that the number of nodes in different classes can be balanced.

Requirements

PyTorch >= 0.4
Python >= 3.6

Usage

Dataset

The datasets can be downloaded from Cora, Citeseer, Pubmed and DBLP. Take Cora dataset as an example:

make a folder called dataset at root directory.
make a folder called cora in dataset directory.
Cora folder contains four files:
- "edge.cora": each line represents a edge between two nodes with the follwing format:
```
1397 1470
1397 362
... ...
```
- "feature.cora": each line represents the features (e.g., one-hot feature) of nodes with the following format:
```
0 0 0 0 1 0 0 0,...,0
0 1 0 1 0 0 0 0,...,0
...
```
- "train.cora": each line represents the node ID in training set with the following format:
```
0
2
...
```
- "test.cora": each line represents the node ID in test set with the following format:
```
2159
2160
...
```

Example Usage

cd ImGAGN
python train.py

Citation

@misc{qu2021imgagnimbalanced,
      title={ImGAGN:Imbalanced Network Embedding via Generative Adversarial Graph Networks}, 
      author={Liang Qu and Huaisheng Zhu and Ruiqi Zheng and Yuhui Shi and Hongzhi Yin},
      year={2021},
      eprint={2106.02817},
      archivePrefix={arXiv},
      primaryClass={cs.LG}
}