StellarGraph Machine Learning Library

Table of Contents

• Introduction
• Guiding Principles
• Getting Started
• Installation
  • Install StellarGraph using pip
  • Install StellarGraph from Github source
• Docker Image
• Running the examples
  • Running the examples with docker
• Algorithms
• Getting Help
• Discourse Community
• CI
• Citing
• References

Introduction

StellarGraph is a Python library for machine learning on graph-structured (or equivalently, network-structured) data.

Graph-structured data represent entities, e.g., people, as nodes (or equivalently, vertices), and relationships between entities, e.g., friendship, as links (or equivalently, edges). Nodes and links may have associated attributes such as age, income, and time when a friendship was established, etc. StellarGraph supports analysis of both homogeneous networks (with nodes and links of one type) and heterogeneous networks (with more than one type of nodes and/or links).

The StellarGraph library implements several state-of-the-art algorithms for applying machine learning methods to discover patterns and answer questions using graph-structured data.

The StellarGraph library can be used to solve tasks using graph-structured data, such as:

• Representation learning for nodes and edges, to be used for visualisation and various downstream machine learning tasks;
• Classification and attribute inference of nodes or edges;
• Link prediction;
• Interpretation of node classification through calculated importances of edges and neighbours for selected nodes [8].

We provide examples of using StellarGraph to solve such tasks using several real-world datasets.

Guiding Principles

StellarGraph uses the Keras library and adheres to the same guiding principles as Keras: user-friendliness, modularity, and easy extendability. Modules and layers of StellarGraph library are designed so that they can be used together with standard Keras layers and modules, if required. This enables flexibility in using existing or creating new models and workflows for machine learning on graphs.

Getting Started

To get started with StellarGraph you'll need data structured as a homogeneous or heterogeneous graph, including attributes for the entities represented as graph nodes. NetworkX is used to represent the graph and Pandas or Numpy are used to store node attributes.

Detailed and narrated examples of various machine learning workflows on network data, supported by StellarGraph, from data ingestion into graph structure to inference, are given in the demos directory of this repository.

Installation

StellarGraph is a Python 3 library and we recommend using Python version 3.6.*. The required Python version can be downloaded and installed from python.org. Alternatively, use the Anaconda Python environment, available from anaconda.com.

Note: while the library works on Python 3.7 it is based on Keras which does not officially support Python 3.7. Therefore, there may be unforseen bugs and you there are many warnings from the Python libraries that StellarGraph depends upon.

The StellarGraph library can be installed in one of two ways, described next.

Install StellarGraph using pip:

To install StellarGraph library from PyPi using pip, execute the following command:

pip install stellargraph

Some of the examples in the demos directory require installing additional dependencies as well as stellargraph. To install these dependencies as well as StellarGraph using pip execute the following command:

pip install stellargraph[demos]

Install StellarGraph from Github source:

First, clone the StellarGraph repository using git:

git clone https://github.com/stellargraph/stellargraph.git

Then, cd to the StellarGraph folder, and install the library by executing the following commands:

cd stellargraph
pip install .

Some of the examples in the demos directory require installing additional dependencies as well as stellargraph. To install these dependencies as well as StellarGraph using pip execute the following command:

pip install .[demos]

Docker Image

• stellargraph/stellargraph: Docker image with stellargraph installed.

Images can be pulled via docker pull stellargraph/stellargraph

Running the examples

See the README in the demos directory for more information about the examples and how to run them.

Algorithms

The StellarGraph library currently includes the following algorithms for graph machine learning:

• GraphSAGE [1]

  • Supports supervised as well as unsupervised representation learning, node classification/regression, and link prediction for homogeneous networks. The current implementation supports multiple aggregation methods, including mean, maxpool, meanpool, and attentional aggregators.

• HinSAGE

  • Extension of GraphSAGE algorithm to heterogeneous networks. Supports representation learning, node classification/regression, and link prediction/regression for heterogeneous graphs. The current implementation supports mean aggregation of neighbour nodes, taking into account their types and the types of links between them.

• attri2vec [4]

  • Supports node representation learning, node classification, and out-of-sample node link prediction for homogeneous graphs with node attributes.

• The Graph ATtention Network (GAT) [5]

  • The GAT algorithm supports representation learning and node classification for homogeneous graphs. There are versions of the graph attention layer that support both sparse and dense adjacency matrices.

• Graph Convolutional Network (GCN) [6]

  • The GCN algorithm supports representation learning and node classification for homogeneous graphs. There are versions of the graph convolutional layer that support both sparse and dense adjacency matrices.

• Simplified Graph Convolutional network (SGC) [7]

  • The SGC network algorithm supports representation learning and node classification for homogeneous graphs. It is an extension of the GCN algorithm that smooths the graph to bring in more distant neighbours of nodes without using multiple layers.

• Node2Vec [2]

  • The Node2Vec and Deepwalk algorithms perform unsupervised representation learning for homogeneous networks, taking into account network structure while ignoring node attributes. The node2vec algorithm is implemented by combining StellarGraph's random walk generator with the word2vec algorithm from Gensim. Learned node representations can be used in downstream machine learning models implemented using Scikit-learn, Keras, Tensorflow or any other Python machine learning library.

• Metapath2Vec [3]

  • The metapath2vec algorithm performs unsupervised, metapath-guided representation learning for heterogeneous networks, taking into account network structure while ignoring node attributes. The implementation combines StellarGraph's metapath-guided random walk generator and Gensim word2vec algorithm. As with node2vec, the learned node representations (node embeddings) can be used in downstream machine learning models to solve tasks such as node classification, link prediction, etc, for heterogeneous networks.

Getting Help

Documentation for StellarGraph can be found here.

Discourse Community

Feel free to ask questions and discuss problems on the StellarGraph Discourse forum.

CI

buildkite integration

Pipeline is defined in .buildkite/pipeline.yml

Docker images

• Tests: Uses the official python:3.6 image.
• Style: Uses black from the stellargraph docker hub organisation.

Citing

StellarGraph is designed, developed and supported by CSIRO's Data61. If you use any part of this library in your research, please cite it using the following BibTex entry

@misc{StellarGraph,
  author = {CSIRO's Data61},
  title = {StellarGraph Machine Learning Library},
  year = {2018},
  publisher = {GitHub},
  journal = {GitHub Repository},
  howpublished = {\url{https://github.com/stellargraph/stellargraph}},
}

References

1. Inductive Representation Learning on Large Graphs. W.L. Hamilton, R. Ying, and J. Leskovec. Neural Information Processing Systems (NIPS), 2017. (link webpage)

2. Node2Vec: Scalable Feature Learning for Networks. A. Grover, J. Leskovec. ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), 2016. (link)

3. Metapath2Vec: Scalable Representation Learning for Heterogeneous Networks. Yuxiao Dong, Nitesh V. Chawla, and Ananthram Swami. ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD), 135–144, 2017 (link)

4. Attributed Network Embedding via Subspace Discovery. D. Zhang, Y. Jie, X. Zhu and C. Zhang, arXiv:1901.04095, [cs.SI], 2019. (link)

5. Graph Attention Networks. P. Velickovic et al. International Conference on Learning Representations (ICLR) 2018 (link)

6. Graph Convolutional Networks (GCN): Semi-Supervised Classification with Graph Convolutional Networks. Thomas N. Kipf, Max Welling. International Conference on Learning Representations (ICLR), 2017 (link)

7. Simplifying Graph Convolutional Networks. F. Wu, T. Zhang, A. H. de Souza, C. Fifty, T. Yu, and K. Q. Weinberger. International Conference on Machine Learning (ICML), 2019. (link)

8. Adversarial Examples on Graph Data: Deep Insights into Attack and Defense. H. Wu, C. Wang, Y. Tyshetskiy, A. Docherty, K. Lu, and L. Zhu. IJCAI 2019. (link)