Difference between model exported from H2O Flow, POJO, MOJO

[gekaprog]

Hi,

I'd like to export a trained model from H2O FLOW for use in Java. What is the difference between 

model exported as binary file, POJO and MOJO? Separately, could you point to an example of training a model in Java?

Thank you

[ang...@0xdata.com]

Hello,

You can see more information about POJO (Plain Old Java Objects) and MOJOs (Model ObJect, Optimized) in the following topics:

- https://github.com/h2oai/h2o-3/blob/master/h2o-docs/src/product/howto/POJO_QuickStart.md
- https://github.com/h2oai/h2o-3/blob/master/h2o-docs/src/product/howto/MOJO_QuickStart.md

Note that POJOs are not supported for source files larger than 1G. Additionally, MOJOs are currently only supported for GBM, DRF, GLM, and K-Means models.

A POJO is downloaded as a single .java file, and a MOJO is downloaded as a zip file containing the domains, trees (for ensemble models), and the model.ini file.

Finally, the following is a link to the Java API guide for the H2O algorithms:
http://docs.h2o.ai/h2o/latest-stable/h2o-algos/javadoc/index.html

Sincerely,
Angela