The Jubatus library is a online machine learning framework which runs in distributed environment.

See http://jubat.us for details.

LGPL 2.1

Improvements

• Simpler interfaces at classifier, regression and recommender

• Clients are NOT COMPATIBLE with previous releases

• Now mix works concurrently in multiple threads (except tf-idf counting)
• Asynchronous RPC to multiple servers at once
• Add --version option
• Interface description language changed from C++-like to Annotated MessagePack-IDL
• Minor error handling
• A bit more tested than previous releases

Bugfix

• #30, #29, #22

Bugfix release: #28

New Features

• recommender

• support fast similar item search, real-time update, distributed data management
• inverted index : exact result, fast search
• locality sensitive hash : approximate result, fast search, small working space

• regression

• online SVR using passive agressive algorithm
• as fast as current classifier

• stat

• a Key(string)-Value(queue<double>)
• O(1) cost of getting sum, standard deviation, max, min, statistic moments for each queue

• server framework

• less-tightly coupled distributed processing framework with each ML implementation
• idl & code generator - make it easy to write own jubatus system
• removed public release of client libraries (so easy to generate!)
• multiple mix - mutiple data objects can be mixed in one jubatus system

Bugfix

• duplicate key entry in fv_converter breaks the parameter

Bugfix release

Hello Jubatus!

First release: including classifier, and mix operation