https://github.com/giannisnik/segk

Learning Structural Node Representations using Graph Kernels
https://github.com/giannisnik/segk

graph-embeddings graph-mining node-embedding representation-learning

Last synced: 2 days ago
JSON representation

Learning Structural Node Representations using Graph Kernels

• Host: GitHub
• URL: https://github.com/giannisnik/segk
• Owner: giannisnik
• License: mit
• Created: 2019-11-26T14:50:29.000Z (almost 5 years ago)
• Default Branch: master
• Last Pushed: 2024-02-02T11:15:51.000Z (10 months ago)
• Last Synced: 2024-08-03T13:14:21.205Z (3 months ago)
• Topics: graph-embeddings, graph-mining, node-embedding, representation-learning
• Language: Jupyter Notebook
• Size: 848 KB
• Stars: 6
• Watchers: 2
• Forks: 1
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
## Learning Structural Node Representations using Graph Kernels

Code for the paper [Learning Structural Node Representations using Graph Kernels](https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8869809).

### Requirements

Code is written in Python 3.6 and requires:

* grakel 0.1b7

* scikit-learn 0.21

### Basic Usage

To run *SEGK* on Barbell graph, execute the following command from the project home directory:


``python segk.py --path-to-edgelist datasets/barbell.edgelist --path-to-output-file embeddings/barbell.txt``

#### Hyperparameters

The following three hyperparameters can be specified:

* radius: the maximum radius of the neighborhood subgraphs

* dim: the dimensionality of the generated embeddings

* kernel: the employed graph kernel (either the shortest path kernel or the weisfeiler lehman kernel)

#### Input

The supported input format is an edgelist where the endpoints are separated by the space character:

    node1_id node2_id

        

#### Output

The output file contains *n* lines (where *n* is the number of nodes of the graph) as follows:

    node_id dim1 dim2 ... dimd

where dim1, ... , dimd is the *d*-dimensional representation learned by *SEGK*.

### Experiments

In the project home directory, there are four jupyter notebooks that reproduce the experiments presented in the paper.

### Cite

Please cite our paper if you use this code:

```

@article{nikolentzos2019learning,

  title={Learning Structural Node Representations using Graph Kernels},

  author={Nikolentzos, Giannis and Vazirgiannis, Michalis},

  journal={IEEE Transactions on Knowledge and Data Engineering},

  year={2019}

}

```