A Python library for machine learning on graphs and networks, offering state-of-the-art algorithms for tasks like node classification and link prediction.
StellarGraph is a Python library for machine learning on graphs and networks, providing state-of-the-art algorithms to analyze graph-structured data. It solves tasks like node classification, link prediction, and representation learning for homogeneous, heterogeneous, and knowledge graphs, helping users uncover patterns in relational data.
Data scientists, machine learning engineers, and researchers working with graph-structured data, such as social networks, knowledge graphs, or biological networks, who need scalable and modular tools for graph machine learning.
Developers choose StellarGraph for its comprehensive suite of graph algorithms, seamless integration with TensorFlow 2 and Keras, and user-friendly design that interoperates with popular Python libraries like Pandas and scikit-learn.
StellarGraph - Machine Learning on Graphs
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Offers a broad range of algorithms from GraphSAGE to TGCN, covering homogeneous, heterogeneous, and temporal graphs as detailed in the algorithms table.
Built on TensorFlow 2 and Keras, enabling easy model construction and training, exemplified in the GCN code snippet.
Includes extensive narrated examples and tutorials, with badges for direct execution in Colab and Binder, as shown in the Getting Started section.
Interoperates smoothly with Pandas, scikit-learn, and other libraries, facilitating data preparation and evaluation in the examples.
Tightly coupled with TensorFlow, making it less flexible for teams invested in PyTorch or other frameworks, and adding significant dependency overhead.
Requires managing multiple dependencies, and some features like igraph support are platform-specific, as noted in the installation section for demos.
Algorithms like GCN use full-batch generators, which can be prohibitive for graphs with millions of nodes without optimizations like Cluster-GCN.