https://github.com/zhengwang100/RECT

This is the source code of "Network Embedding with Completely-Imbalanced Labels". TKDE2020
https://github.com/zhengwang100/RECT

graph-embedding graph-representation-learning network-embedding

Last synced: about 1 month ago
JSON representation

This is the source code of "Network Embedding with Completely-Imbalanced Labels". TKDE2020

• Host: GitHub
• URL: https://github.com/zhengwang100/RECT
• Owner: zhengwang100
• Created: 2020-02-23T07:23:35.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2021-03-23T06:05:58.000Z (over 3 years ago)
• Last Synced: 2024-08-01T22:41:47.797Z (4 months ago)
• Topics: graph-embedding, graph-representation-learning, network-embedding
• Language: Python
• Size: 9.67 MB
• Stars: 18
• Watchers: 5
• Forks: 7
• Open Issues: 0
• Metadata Files:
  • Readme: readme.md

Awesome Lists containing this project

• awesome-network-embedding - [Python

README

        
# RECT (python source code)

Network Embedding with Completely-imbalanced Labels. TKDE2020 [paper](https://zhengwang100.github.io/pdf/TKDE20_wzheng.pdf).

This is a deep method for the problem of [Zero-shot Graph Embedding (ZGE)](https://zhengwang100.github.io/project/zero_shot_graph_embedding.html), i.e., graph embeddings when labeled data cannot cover all classes. 

Breifly, RECT contains two parts:

---

- RECT-L is the supervised part in which a semantic loss is used. 

- RECT-N is the unsupervised part in which the network structure is preserved. Note, this part can be replaced by any unsupervised NRL methods.

Usage (abstract):

---

- set the dataset 

- python main_rect.py

```

------ evaluate RECT-N ---------

Training an SVM classifier under the pre-defined split setting...

(0.7335058214747736, 0.670830503861163)

------ evaluate RECT-L ---------

Training an SVM classifier under the pre-defined split setting...

(0.7141871496334627, 0.6402691559469643)

------ evaluate RECT ---------

Training an SVM classifier under the pre-defined split setting...

(0.7441138421733506, 0.6805281849343917)

```

Citing

---

If you find this useful in your research, please cit our paper, thx:

```

@article{wang2020RECT,

  title={Network Embedding with Completely-imbalanced Labels},

  author={Wang, Zheng and Ye, Xiaojun and Wang, Chaokun and Cui, Jian and Yu, Philip S},

  journal={TKDE},

  year={2020},

  doi = {10.1109/TKDE.2020.2971490},

  publisher={IEEE}

}

```