boostsrl 0.4.1

Python wrappers for using BoostSRL jar files.

boostsrl is a set of Python wrappers around BoostSRL with a scikit-learn interface.

Questions? Contact Alexander L. Hayes (hayesall)

Getting Started

Prerequisites:

• Java 1.8

• Python (3.6, 3.7)

Installation

pip install boostsrl

Basic Usage

The general setup should be similar to scikit-learn. But there are a few extra requirements in terms of setting background knowledge and formatting the data.

A minimal working example (using the Toy-Cancer data set imported with ‘example_data’) is:

>>> from boostsrl.rdn import RDN
>>> from boostsrl import Background
>>> from boostsrl import example_data
>>> bk = Background(
...     modes=example_data.train.modes,
... )
>>> clf = RDN(
...     background=bk,
...     target='cancer',
...     max_tree_depth=2,
...     node_size=1,
... )
>>> clf.fit(example_data.train)
>>> clf.predict_proba(example_data.test)
array([0.88079619, 0.88079619, 0.88079619, 0.3075821 , 0.3075821 ])
>>> clf.classes_
array([1., 1., 1., 0., 0.])

example_data.train and example_data.test are each boostsrl.Database objects, so this hides some of the complexity behind the scenes.

This example abstracts away some complexity in exchange for compactness. For more thorough examples, see the ‘docs/examples/’ directory.

Contributing

Please refer to CONTRIBUTING.md for information on submitting issues and pull requests.

Versioning

We use SemVer for versioning. See Releases for stable versions that are available, or the Project Page on PyPi.