This repository contains a Python implementation of the DeepWalk algorithm, including the Skip-gram model. DeepWalk is a popular algorithm for learning latent representations of vertices in a network, which can be used for various downstream tasks such as node classification, link prediction, and community detection.
- Implementation of the DeepWalk algorithm based on the karateclub library by Benedek Rozemberczki
- Skip-gram model for learning node embeddings
- Random walk generator for sampling node sequences from the graph
- Training loop with early stopping mechanism
- Visualization of learned node embeddings using t-SNE dimensionality reduction
- Python 3.x
- NumPy
- NetworkX
- scikit-learn
- Matplotlib
-
Clone the repository:
git clone https://github.com/your-username/deepwalk-implementation.git -
Install the required dependencies:
pip install numpy networkx scikit-learn matplotlib -
Run the
main.ipynbnotebook to train the DeepWalk model on the Zachary's Karate Club dataset and visualize the learned node embeddings.
activation.py: Contains the softmax activation function.deepwalk.py: Implements the DeepWalk algorithm and the random walk generator.loss.py: Defines the cross-entropy loss function.main.ipynb: Notebook demonstrating the usage of the DeepWalk implementation on the Karate Club dataset.skip_gram.py: Implements the Skip-gram model for learning node embeddings.
- The current implementation does not include hierarchical softmax or negative sampling, which are techniques used in the original DeepWalk paper to improve training efficiency.
- Perozzi, B., Al-Rfou, R., & Skiena, S. (2014). DeepWalk: Online Learning of Social Representations. In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining (pp. 701-710).
- karateclub library by Benedek Rozemberczki
This project is licensed under the MIT License. See the LICENSE file for details.