The release version of Hivemall v0.4.0

This is the stable release of Hivemall v0.4.0.

This version makes major development leaps and includes lots of changes. Major enhancements in this release includes supports for Factorization Machine (usage 1) and RandomForest (usage 1, 2).

Last but not least, I would like to thank contributors who made contributions to this release.

Changes since v0.3.2-3 are summarized as follows:

• Major Enhancement
  • Introduced RandomForest classifier/regressor using Smile #219
  • Introduced Factorization Machine classifier/regressor #207
• Minor Enhancement
  • Added quantify UDTF which is useful for assigning numbers to non-numeric features. [c51fbb4]
  • Added quantified_features UDF which is used for creating sparse quantitative features [22d8897]
  • Added normalize_unicode UDF [1ca0aac]
  • Added feature_index UDF [cc21ec9]
  • Added quantitative_features UDF which is use for creating a sparse quantitative features [3ce8e9f]
  • Implemented gradient_tree_boosting_classifier (still in Beta) [0e61260]
  • Added guess_attribute_types UDF. [54dfc23]
• Major Changes
  • Changed behavior of categorical_features UDF to always makes categorial features [d6f84f2]
  • Changed behavior of vectorize_features to parse numbers as double instead of float [982e079]
  • Fixed to include Netty jars to hivemall-fat.jar [6685d75]
• Minor Changes
  • Added "-help" option to UDTF that shows an usage of the function [9603460]
  • Added -workers option to MixServer #212
  • Added dependency to log4j to hivemall-fat.jar [4b98bae]
  • Removed ant build file. Use Maven instead. [568ef5d]
• Bugfix
  • Fixed a bug in tf UDAF [224057b]
  • Fixed a bug in diffclock computation logic in MixServer #220
  • Fixed a bug for sigmoid(null). Treasure Data PLT-4718. [ffe213a]
  • Fixed a bug in normalization functions that results becomes NaN when divided by zero [624e375]