Product and service reviews are conducted independently by our editorial team, but we sometimes make money when you click on links. Learn more.

Machine Learning Platform H2O Version 3.0 Software Released

By - Source: Toms IT Pro released version 3.0 of the H2O open source machine learning platform. The new version presents data through an updated web user interface (UI) and brings more usability through workflows. Data connections from other services will be available through a more robust API, and H2O 3.0 can now be integrated with Python and Sparkling Water. Sparkling Water is an application that brings H2O machine learning software onto Apache Spark clusters.

Sri Ambati, CEO and co-founder of H2O, said version 3.0 is a machine learning platform that will appeal to developers trying to make use of the swell of data that the Internet of Things will bring. "Prediction is the new search," said Ambati. "Our APIs and embeddable workflows will become the brain to power the apps in the Internet of sensors through the cloud."

The Python API for H2O version 3 allows developers to work with H2O through the Python command line or through integrated development environments such as iPython Notebook. The Python API allows direct access to the data held in H2O clusters.

Behind the scenes, the Python API is interacting with a REST API which is also powering the H2O's Flow UI. Flow UI is a web-based "control panel" that lets you import your data, create data models, and view predictions and analysis all with command-line tools and point-and-click data view manipulations.

Apache Spark developers can take advantage of H2O machine learning through the Sparkling Water application. Sparkling Water allows interactive use through the Scala shell and access to H2O through the Spark APIs. Users running H2O on Spark now get all of the REST API management tools, including Flow UI and Python.

H2O version 3.0 includes features to help improve the overall usability and make it more accessible to new users. Flow UI has new visualizations and more point-and-click options for doing more with datasets through a natural interface. There are more interactive tutorials. There is also an "Assist Me!" button that teaches new users how to perform common tasks.

Performance in version 3.0 is improved, as well. K-Means, Generalized Linear Models, Distributed Random Forest, Naïve Bayes, Gradient Boosted Machines, Deep Learning, Quantiles and Anomaly Detection algorithms all show improved performance in the new H2O.

The new H2O version 3.0 is available now at