Metadata-Version: 2.1
Name: gtrick
Version: 0.0.dev1
Summary: Bag of Tricks for Graph Neural Networks
Home-page: https://github.com/sangyx/gtrick
Author: sangyx
Author-email: sangyunxin@gmail.com
License: UNKNOWN
Platform: UNKNOWN
Classifier: Programming Language :: Python :: 3
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: >=3
Description-Content-Type: text/markdown
License-File: LICENSE

# gtrick: Bag of Tricks for Graph Neural Networks

> Trick is all you need.

## Trick

|     Trick    | Example | Task | Reference |
|:------------:|:------------:|:------------:|:-----:|
| Virtual Node |  [DGL](https://nbviewer.org/github/sangyx/gtrick/blob/main/benchmark/dgl/VirtualNode.ipynb)<br>[PyG](https://nbviewer.org/github/sangyx/gtrick/blob/main/benchmark/pyg/VirtualNode.ipynb) | graph | [Neural Message Passing for Quantum Chemistry](https://arxiv.org/pdf/1704.01212.pdf) |
| FLAG |  [DGL](https://nbviewer.org/github/sangyx/gtrick/blob/main/benchmark/dgl/FLAG.ipynb)<br>[PyG](https://nbviewer.org/github/sangyx/gtrick/blob/main/benchmark/pyg/FLAG.ipynb) | node*<br>graph | [Robust Optimization as Data Augmentation for Large-scale Graphs](https://arxiv.org/abs/2010.09891) |

## Benchmark

The results listed below are implemented by PyG. You can find the results of DGL in [DGL Benchmark](benchmark/dgl/README.md).

### Graph Property Prediction: ogbg-molhiv

|     Trick     |       GCN       |       GIN       |
|:-------------:|:---------------:|:---------------:|
|       —       | 0.7690 ± 0.0053 | 0.7778 ± 0.0130 |
| +Virtual Node | 0.7581 ± 0.0135 | 0.7713 ± 0.0036 |
|     +FLAG     | 0.7627 ± 0.0124 | 0.7764 ± 0.0083 |

### Node Property Prediction: ogbn-arxiv
|     Trick     |       GCN       |       SAGE      |
|:-------------:|:---------------:|:---------------:|
|       —       | 0.7152 ± 0.0024 | 0.7153 ± 0.0028 |
|     +FLAG     | 0.7187 ± 0.0020 | 0.7206 ± 0.0013 |

