Update README.md
Browse files
README.md
CHANGED
@@ -9,4 +9,55 @@ language:
|
|
9 |
- en
|
10 |
datasets:
|
11 |
format: csv
|
12 |
-
---
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
9 |
- en
|
10 |
datasets:
|
11 |
format: csv
|
12 |
+
---
|
13 |
+
|
14 |
+
|
15 |
+
---
|
16 |
+
license: apache-2.0
|
17 |
+
tags:
|
18 |
+
- text
|
19 |
+
- graph
|
20 |
+
task_categories:
|
21 |
+
- graph-ml
|
22 |
+
language:
|
23 |
+
- en
|
24 |
+
datasets:
|
25 |
+
format: csv
|
26 |
+
---
|
27 |
+
|
28 |
+
The dataset is dynamic graphs for paper [CrossLink](https://arxiv.org/pdf/2402.02168.pdf). The usage of this dataset can be seen in [Github](https://weichow23.github.io/CrossLink/)
|
29 |
+
|
30 |
+
## 🚀 Introduction
|
31 |
+
|
32 |
+
CrossLink learns the evolution pattern of a specific downstream graph and subsequently makes pattern-specific link predictions.
|
33 |
+
It employs a technique called *conditioned link generation*, which integrates both evolution and structure modeling to perform evolution-specific link prediction. This conditioned link generation is carried out by a transformer-decoder architecture, enabling efficient parallel training and inference. CrossLink is trained on extensive dynamic graphs across diverse domains, encompassing 6 million dynamic edges. Extensive experiments on eight untrained graphs demonstrate that CrossLink achieves state-of-the-art performance in cross-domain link prediction. Compared to advanced baselines under the same settings, CrossLink shows an average improvement of **11.40%** in Average Precision across eight graphs. Impressively, it surpasses the fully supervised performance of 8 advanced baselines on 6 untrained graphs.
|
34 |
+
|
35 |
+
![Architecture](model.png)
|
36 |
+
|
37 |
+
## Format
|
38 |
+
|
39 |
+
Please keep the dataset in the fellow format:
|
40 |
+
|
41 |
+
| Unnamed: 0 | u | i | ts | label | idx |
|
42 |
+
| ---------- | ------------- | ------------- | ------------------ | ------------ | ---------------------- |
|
43 |
+
| `idx-1` | `source node` | `target node` | `interaction time` | `defalut: 0` | `from 1 to the #edges` |
|
44 |
+
|
45 |
+
You can prepare those data by the code in `preprocess_data` folder
|
46 |
+
|
47 |
+
You can also use our processed data in [huggingface](https://huggingface.co/datasets/WeiChow/DyGraphs)
|
48 |
+
|
49 |
+
## 📚 Citation
|
50 |
+
|
51 |
+
If you find this work helpful, please consider citing:
|
52 |
+
|
53 |
+
```bibtex
|
54 |
+
@misc{huang2024graphmodelcrossdomaindynamic,
|
55 |
+
title={One Graph Model for Cross-domain Dynamic Link Prediction},
|
56 |
+
author={Xuanwen Huang and Wei Chow and Yang Wang and Ziwei Chai and Chunping Wang and Lei Chen and Yang Yang},
|
57 |
+
year={2024},
|
58 |
+
eprint={2402.02168},
|
59 |
+
archivePrefix={arXiv},
|
60 |
+
primaryClass={cs.LG},
|
61 |
+
url={https://arxiv.org/abs/2402.02168},
|
62 |
+
}
|
63 |
+
```
|