New versions of H2O-3 and Sparkling Water available

Key takeaways

The post announces new releases of H2O-3.16 and Sparkling Water 2.0, 2.1, and 2.2 ahead of that year's H2O World conference.
The post lists H2O-3.16 features including MOJO support for Stacked Ensembles, a selectable meta-learner algorithm, and cross-validation metrics as the new default for the AutoML leaderboard.
The post notes that AutoML gained multiclass stacking support in this release, removing a prior check that skipped stacking for multiclass problems.
The post lists Sparkling Water 2.0 through 2.2 improvements, including support for H2O models in Spark Python pipelines and improved handling of sparse vectors in the internal cluster.

Dear H2O Community, #H2OWorld is on Monday and we can’t wait to see you there! We’ll also be live streaming the event starting at 9:25am PST. Explore the agenda here . Today we’re excited to share that new versions of H2O-3 and Sparkling Water are available. We invite you to download them here: http://www.h2o.ai/download/

H2O-3.16 – MOJOs are now supported for Stacked Ensembles. – Easily specify the meta-learner algorithm type that Stacked Ensemble should use. This can be AUTO, GLM, GBM, DRF or Deep Learning . – GBM, DRF now support custom evaluation metrics. – The AutoML leaderboard now uses cross-validation metrics (new default). – Multiclass stacking is now supported in AutoML . Removed the check that caused AutoML to skip stacking for multiclass. – The Aggregator Function is now exposed in the Python/R client. – Support for Python 3.6. Detailed changes and bug fixes can be found here: https://github.com/h2oai/h2o-3/blob/master/Changes.md Sparkling Water 2.0, 2.1, 2.2 – Support for H2O Models into Spark python pipelines. – Improved handling of sparse vectors in internal cluster. – Improved stability of external cluster deployment mode. – Includes latest H2O-3.16.0.2. Detailed changes and bug fixes can be explored here: 2.2 – https://github.com/h2oai/sparkling-water/blob/rel-2.2/doc/CHANGELOG.rst 2.1 – https://github.com/h2oai/sparkling-water/blob/rel-2.1/doc/CHANGELOG.rst 2.0 – https://github.com/h2oai/sparkling-water/blob/rel-2.0/doc/CHANGELOG.rst Hope to see you on Monday! The H2O.ai Team

Key takeaways

The post announces new releases of H2O-3.16 and Sparkling Water 2.0, 2.1, and 2.2 ahead of that year's H2O World conference.
The post lists H2O-3.16 features including MOJO support for Stacked Ensembles, a selectable meta-learner algorithm, and cross-validation metrics as the new default for the AutoML leaderboard.
The post notes that AutoML gained multiclass stacking support in this release, removing a prior check that skipped stacking for multiclass problems.
The post lists Sparkling Water 2.0 through 2.2 improvements, including support for H2O models in Spark Python pipelines and improved handling of sparse vectors in the internal cluster.