The StellarGraph team is happy to announce the release of StellarGraph version 0.11.
StellarGraph is an open-source library featuring state-of-the-art graph machine learning algorithms. The project is delivered as part of CSIRO’s Data61.
Five algorithms and streamlined onboarding are only some of the major new features delivered in the 0.11 release of the library, with substantial API, documentation and demo improvements to support users in running and exploring StellarGraph functionality.
New algorithms include:
- Watch Your Step: computes node embeddings by simulating the effect of random walks, rather than doing them.
- Deep Graph Infomax: performs unsupervised node representation learning.
- Temporal Random Walks (Continuous-Time Dynamic Network Embeddings): random walks that respect the time that each edge occurred.
- ComplEx: computes multiplicative complex-number embeddings for entities and relationships (edge types) in knowledge graphs, used for link prediction.
- DistMult: computes multiplicative real-number embeddings for entities and relationships (edge types) in knowledge graphs, used for link prediction.
A sixth algorithm remains under active development, but is available in this release as an experimental preview:
- GCNSupervisedGraphClassification: supervised graph classification model based on Graph Convolutional layers (GCN).
The new release incorporates extensive enhancements and bug fixes, some of which include:
- StellarGraph.to_adjacency_matrix is at least 15 times faster on undirected graphs
- ClusterNodeGenerator is noticeably faster, reducing the time taken to train and predict with a ClusterGCN model
- Addition of a subgraph method for computing a node-induced subgraph
- Addition of a connected_components method for computing the nodes involved in each connected component in a StellarGraph
- The info method has been improved for heterogeneous graphs with many types and also shows information about the size and type of each node type’s feature vectors
- Four new datasets added to stellargraph.datasets
- Neo4j functionality is now tested on CI
- Example Jupyter notebooks can now be run directly in Google Colab and Binder
- New notebooks demonstrating how to construct a StellarGraph object from Pandas, and from NetworkX
See the full release notes here.