A parallel implementation of "graph2vec: Learning Distributed Representations of Graphs" (MLGWorkshop 2017).
Alternatives To Graph2vec
Project NameStarsDownloadsRepos Using ThisPackages Using ThisMost Recent CommitTotal ReleasesLatest ReleaseOpen IssuesLicenseLanguage
Gun17,5392461639 days ago236December 23, 2022301otherJavaScript
An open source cybersecurity protocol for syncing decentralized graph data.
Tflearn9,602481822 days ago6November 11, 2020576otherPython
Deep learning library featuring a higher-level API for TensorFlow.
a month ago644gpl-3.0Python
Translate darknet to tensorflow. Load trained weights, retrain/fine-tune using tensorflow, export constant graph def to mobile devices
Smile5,83312135a month ago33June 14, 202312otherJava
Statistical Machine Intelligence & Learning Engine
2 years ago217otherShell
TensorFlow for macOS 11.0+ accelerated using Apple's ML Compute framework.
Stellargraph2,768263 months ago24February 22, 2021318apache-2.0Python
StellarGraph - Machine Learning on Graphs
Yolov3 Tf22,501
2 months ago167mitJupyter Notebook
YoloV3 Implemented in Tensorflow 2.0
4 years ago4mit
A comprehensive collection of recent papers on graph deep learning
8 months ago61bsd-3-clausePython
Deep Learning toolkit for Computer Vision.
Pytextrank2,057611a month ago19August 07, 202312mitPython
Python implementation of TextRank algorithms ("textgraphs") for phrase extraction
Alternatives To Graph2vec
Select To Compare

Alternative Project Comparisons


Arxiv codebeat badge repo size benedekrozemberczki


Recent works on representation learning for graph structured data predominantly focus on learning distributed representations of graph substructures such as nodes and subgraphs. However, many graph analytics tasks such as graph classification and clustering require representing entire graphs as fixed length feature vectors. While the aforementioned approaches are naturally unequipped to learn such representations, graph kernels remain as the most effective way of obtaining them. However, these graph kernels use handcrafted features (e.g., shortest paths, graphlets, etc.) and hence are hampered by problems such as poor generalization. To address this limitation, in this work, we propose a neural embedding framework named graph2vec to learn data-driven distributed representations of arbitrary sized graphs. graph2vec's embeddings are learnt in an unsupervised manner and are task agnostic. Hence, they could be used for any downstream task such as graph classification, clustering and even seeding supervised representation learning approaches. Our experiments on several benchmark and large real-world datasets show that graph2vec achieves significant improvements in classification and clustering accuracies over substructure representation learning approaches and are competitive with state-of-the-art graph kernels.

The model is now also available in the Karate Club package.

This repository provides an implementation for graph2vec as it is described in:

graph2vec: Learning distributed representations of graphs. Narayanan, Annamalai and Chandramohan, Mahinthan and Venkatesan, Rajasekar and Chen, Lihui and Liu, Yang MLG 2017, 13th International Workshop on Mining and Learning with Graphs (MLGWorkshop 2017).

The original TensorFlow implementation is available [here].


The codebase is implemented in Python 3.5.2 | Anaconda 4.2.0 (64-bit). Package versions used for development are just below.

jsonschema        2.6.0
tqdm              4.28.1
numpy             1.15.4
pandas            0.23.4
texttable         1.5.0
gensim            3.6.0
networkx          2.4
joblib            0.13.0


The code takes an input folder with json files. Every file is a graph and files have a numeric index as a name. The json files have two keys. The first key called "edges" corresponds to the edge list of the graph. The second key "features" corresponds to the node features. If the second key is not present the WL machine defaults to use the node degree as a feature. A sample graph dataset from NCI1 is included in the `dataset/` directory.


Learning of the embedding is handled by the src/ script which provides the following command line arguments.

Input and output options

  --input-path   STR    Input folder.           Default is `dataset/`.
  --output-path  STR    Embeddings path.        Default is `features/nci1.csv`.

Model options

  --dimensions     INT          Number of dimensions.                             Default is 128.
  --workers        INT          Number of workers.                                Default is 4.
  --epochs         INT          Number of training epochs.                        Default is 1.
  --min-count      INT          Minimal feature count to keep.                    Default is 5.
  --wl-iterations  INT          Number of feature extraction recursions.          Default is 2.
  --learning-rate  FLOAT        Initial learning rate.                            Default is 0.025.
  --down-sampling  FLOAT        Down sampling rate for frequent features.         Default is 0.0001.


The following commands learn an embedding of the graphs and writes it to disk. The node representations are ordered by the ID. Creating a graph2vec embedding of the default dataset with the default hyperparameter settings. Saving the embedding at the default path.

$ python src/

Creating an embedding of an other dataset. Saving the output in a custom place.

$ python src/ --input-path new_data/ --output-path features/nci2.csv

Creating an embedding of the default dataset in 32 dimensions.

$ python src/ --dimensions 32


Popular Graph Projects
Popular Machine Learning Projects
Popular Computer Science Categories
Related Searches

Get A Weekly Email With Trending Projects For These Categories
No Spam. Unsubscribe easily at any time.
Machine Learning
Unsupervised Learning
Matrix Factorization