Graph Data Science Library v1.5 now available on github, our download center, and desktop 
This release includes:
- ML based link prediction and node classification (train a machine learning model in Neo4j; includes appropriate graph splitting and model selection to save only the best model) with
gds.alpha.ml.nodeClassificationandgds.alpha.ml.linkPrediction - Persistent & publishable models (train graphsage and persist your model through a db restart!) (Enterprise only)
- New algos: speaker listener label propagation & HITS
- New in memory graph compression format for Enterprise Edition users (up to 75% less memory!)
- Export from the in-memory graph to CSV
- Pathfinding algos (Dijkstra, Yen's, A*) reimplemented and moved to the
betatier
Plus other bug fixes and improvements.
Full details are available in the release notes, and the docs are available at https://neo4j.com/docs/graph-data-science/1.5/.
We've got a webinar scheduled for March 11th to walk through the new features and do a demo, sign up here