Helper functions

Mostly written in Java, these functions build search indexes, compute dataset characteristics, generate tasks and evaluate the results of certain tasks.

Code

The Java code is available in the 'OpenML' repository: https://github.com/openml/OpenML/tree/master/Java

Components

General:

• OpenML: Building Lucene search index and smaller tools, e.g. extracting documentation from WEKA source files and ARFF files
• generateApiDocs: Generates API HTML Documentation
• http_post_file: Example how to post files to the api using Java.

Support for tasks:

• foldgeneration: Java code for generating cross-validation folds. Can be used from command line.
• splitgeneration: Split generator for cross validation and holdout. Unsure what's the difference with the previous?
• generate_predictions: Helper class to build prediction files based on WEKA output. Move to WEKA repository?
• evaluate_predictions: The evaluation engine computing evaluation scores based on submitted predictions