Scikit-learn estimators based on projection pursuit.
This repository is home to a couple scikit-learn-compatible estimators based on Jerome Friedman's generalizations of his and Werner Stuetzle's Projection Pursuit Regression algorithm. A regressor capable of multivariate estimation and dimensionality reduction and a univariate classifier based on regression to a one-hot multivariate representation are included.
This repository is also meant to serve as a fairly pared-down example of how to use TravisCI, Coveralls, Sphinx, PyTest, how to deploy to PyPI and Github Pages, and how to create a Scikit-Learn Estimator that passes the sklearn checks and follows the PEP 8 style standard.
Installation and Usage
The package by itself comes with a single module containing the estimators. Before
installing the module you will need
To install the module execute:
pip install projection-pursuit
$ python setup.py install
If the installation is successful, you should be able to execute the following in Python:
>>> from skpp import ProjectionPursuitRegressor >>> estimator = ProjectionPursuitRegressor() >>> estimator.fit(np.arange(10).reshape(10, 1), np.arange(10))
Sphinx is run via continuous integration to generate the API.
For a few usage examples, see the examples and benchmarks directories. For an intuition of what the learner is doing, try running
viz_training_process.py. For comparisons to other learners and an intuition of why you might want to try PPR, try the benchmarks. For a deep dive in to the math and an explanation of exactly how and why this works, see
- Friedman, Jerome. (1985). "Classification and Multiple Regression Through Projection Pursuit." http://www.slac.stanford.edu/pubs/slacpubs/3750/slac-pub-3824.pdf
- Hastie, Tibshirani, & Friedman. (2016). The Elements of Statistical Learning 2nd Ed., section 11.2.
- (2017) Projection pursuit regression https://en.wikipedia.org/wiki/Projection_pursuit_regression
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.