Skip to content

Neo4j Graph Data Science library preview release 0.9.0

Compare
Choose a tag to compare
@jjaderberg jjaderberg released this 07 Feb 13:00
· 21737 commits to master since this release

neo4j-graph-data-science-0.9.0

Breaking changes

  • Product-supported algorithms (WCC, Louvain, Label Propagation, Page Rank and Node Similarity) have been ported to the new procedure API.
  • Removed beta versions of graph catalog procedures
  • Weighted algorithms now run on an unweighted graph copy unless a relationship weight is specified in the procedure configuration.

New features

  • Added new catalog procedures:
    • gds.graph.create -- adds a graph to the catalog
    • gds.graph.create.cypher -- adds a graph to the catalog using Cypher
    • gds.graph.list -- lists information about graphs in the catalog
    • gds.graph.exists -- checks if a graph exists in the catalog
    • gds.graph.drop -- drops a graph from the catalog
    • gds.graph.create.estimate -- estimate memory usage for a graph
    • gds.graph.create.cypher.estimate -- estimate memory usage for a graph using Cypher
  • Added new catalog function:
    • gds.graph.exists -- checks if a graph exists in the catalog
  • Added support for multiple relationship types and properties in Cypher projection.
  • For Cypher projections, node properties are initialized from the RETURN statement if no nodeProperties are given.

Bug fixes

  • Fixed potential assertion error when running memory estimation on pre-sized graph.
  • Fixed bug in Node Similarity where it would return wrong node ids.
  • Fixed failure when assigning integer values to parameters that expect doubles in procedures.

neo4j-graph-data-science-0.9.0-standalone.jar

This jar contains the Neo4j Graph Data Science Library plugin.
To use it in your Neo4j database, copy the jar into the plugins directory of your Neo4j 3.5 installation and restart the server.
For more information visit https://neo4j.com/docs/graph-data-science/preview/