AM-GCN - GitLite

KO | EN

gitlite — search

#javascript #android #machine-learning #rust #nodejs #java #windows #awesome #kubernetes #awesome-list #pytorch #security #mcp #openai

AM-GCN

(★ 261)

AM-GCN: Adaptive Multi-channel Graph Convolutional Networks

파일 탐색기

최종 버전 다운로드 (.zip)

- - config.cpython-37.pyc
  - layers.cpython-36.pyc
  - layers.cpython-37.pyc
  - models.cpython-36.pyc
  - models.cpython-37.pyc
  - utils.cpython-36.pyc
  - utils.cpython-37.pyc
- - Case1.py
  - Case2.py
- - 20acm.ini
  - 20BlogCatalog.ini
  - 20citeseer.ini
  - 20coraml.ini
  - 20flickr.ini
  - 20uai.ini
  - 40acm.ini
  - 40BlogCatalog.ini
  - 40citeseer.ini
  - 40coraml.ini
  - 40flickr.ini
  - 40uai.ini
  - 60acm.ini
  - 60BlogCatalog.ini
  - 60citeseer.ini
  - 60coraml.ini
  - 60flickr.ini
  - 60uai.ini
- __init__.py
- config.py
- dataprocess.py
- layers.py
- main.py
- models.py
- utils.py
- acm.zip
- BlogCatalog.zip
- citeseer.zip
- coraml.zip
- flickr.zip
- uai.zip
LICENSE
README.md

// repository documentation

Was this content helpful?

★ 0 (0 ratings)

# AM-GCN
Source code for KDD2020 "[AM-GCN: Adaptive Multi-channel Graph Convolutional Networks](http://shichuan.org/doc/86.pdf)"

# Environment Settings 
* python == 3.7   
* Pytorch == 1.1.0  
* Numpy == 1.16.2  
* SciPy == 1.3.1  
* Networkx == 2.4  
* scikit-learn == 0.21.3

# Usage 
````
python main.py -d dataset -l labelrate
````
* **dataset**: including \[citeseer, uai, acm, BlogCatalog, flickr, coraml\], required.  
* **labelrate**: including \[20, 40, 60\], required.

e.g.  
````
python main.py -d citeseer -l 20
````

# Data
## Link
* **Citeseer**: [Semi-Supervised Classifcation with Graph Convolutional Networks.](https://github.com/tkipf/gcn)  
* **UAI2010**: A Unifed Weakly Supervised Framework for Community Detection and Semantic Matching. 
* **ACM**: [Heterogeneous Graph Attention Network.](https://github.com/Jhy1993/HAN)  
* **BlogCatalog,Flickr**: [Co-Embedding Attributed Networks.](https://github.com/mengzaiqiao/CAN)  
* **CoraFull**: [Deep Gaussian Embedding of Graphs: Unsupervised Inductive Learning via Ranking.](https://github.com/abojchevski/graph2gauss/)

## Usage
Please first **unzip** the data folders and then use. The files in folders are as follows:
````
citeseer/
├─citeseer.edge: edge file.  
├─citeseer.feature: feature file.  
├─citeseer.label: label file.  
├─testL/C.txt: test file. L/C, i.e., Label pre Class, L/C = 20, 40, 60.   
├─trainL/C.txt: train file. L/C, i.e., Label pre Class, L/C = 20, 40, 60.  
└─knn
   └─ck.txt: feature graph file. k = 2~9
````
# Parameter Settings

Recorded in   **./AMGCN/config/[L/C][dataset].ini**  
e.g.   **./AMGCN/config/20citeseer.ini**

* **Model_setup**: parameters for training AM-GCN, such as nhid1, nhid2, beta, theta... 
* **Data_setting**: dataset setttings, such as paths for input, node numbers, feature dimensions...

# Reference
````
@inproceedings{wang2020gcn,
  title={AM-GCN: Adaptive Multi-channel Graph Convolutional Networks},
  author={Wang, Xiao and Zhu, Meiqi and Bo, Deyu and Cui, Peng and Shi, Chuan and Pei, Jian},
  booktitle={Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery \& Data Mining},
  pages={1243--1253},
  year={2020}
}
````