oll 0.1

Online machine learning library (based on OLL C++ library)

oll-python

This is a Python binding of the OLL 0.03 library for machine learning. OLL is specialized for large-scale, but sparse, learning tasks such as Natural Language Processing.

Currently, OLL 0.03 supports following online-learning algorithms:

• Perceptron

• Averaged Perceptron

• Passive Agressive (PA, PA-I, PA-II)

• ALMA (modified slightly from original)

• Confidence Weighted Linear-Classification.

While these algorithms are very efficient in terms of speed, and space (linear in the number of examples, and features), its performances are comparable to the batch-style learning methods such as SVMs, MEs.

For details of OLL, see: http://code.google.com/p/oll

Installation

$ pip install oll

OLL library is bundled, so you don’t need to install it separately.

Usage

>>> import oll
>>> o = oll.oll("CW")
>>> o.add({0: 1.0, 1: 2.0, 2: -1.0}, 1) # train
>>> o.classify({0:1.0, 1:1.0}) # predict
>>> o.save('oll.model')
>>> o.load('oll.model')

Note

• This module requires C++ compiler to build.

• oll.cpp & oll.hpp : Copyright (c) 2011, Daisuke Okanohara

• oll_swig_wrap.cxx is generated based on ‘oll_swig.i’ in oll-ruby (https://github.com/syou6162/oll-ruby)

License

New BSD License.

CHANGES

0.1 (2013-10-11)

Initial release.