Welcome to sklearn-extensions’s documentation!

Scikit-Learn Extensions (sklearn_extensions) is a single source repository for extensions to [scikit-learn](https://github.com/sklearn/sklearn). It is intended to compliment the slower more cautious approach of scikit-learn with regard to adding new predictors and modules, with a separate pip-installable source for sklearn-compatible modules that may not meet those standards.

In particular, this project is interested in smaller one-off projects, particularly even gists, rather than larger more established ones (such as pylearn2, lifeline, or lightning). Other than larger projects, we will shy away from projects with significant external dependencies (i.e. wrappers around vowpal wabbit or xgboost), and rather prefer more python/numpy/scipy based projects.

Due to these guiding goals, the modules included here may not be as well tested, production ready, or stable as those included directly in sklearn. This is pretty much the wild west, test anything that uses this package heavily.

Installation / Usage

We aim to first support python 3, and are hosted on pypi, so to install just:

pip install sklearn-extensions

Note that the install here will install all underlying packages, and is therefore pretty big. It is recommended that you do this in a virtualenv.

Extensions Included So Far

  • Kernel Regression
    • kernel_regression.KernelRegression()
  • Fuzzy K-Means
    • fuzzy_kmeans.KMeans()
    • fuzzy_kmeans.KMedians()
    • fuzzy_kmeans.FuzzyKMeans()
  • Sparse Filtering
    • sparse_filtering.SparseFiltering()
  • Extreme Learning Machines
    • extreme_learning_machines.ELMRegressor()
    • extreme_learning_machines.ELMClassifier()
    • extreme_learning_machines.GenELMRegressor()
    • extreme_learning_machines.GenELMClassifier()
    • extreme_learning_machines.RandomLayer()
    • extreme_learning_machines.GRBFRandomLayer()
    • extreme_learning_machines.RBFRandomLayer()
    • extreme_learning_machines.MLPRandomLayer()
  • Non-negative Garotte
    • non_negative_garotte.NonNegativeGarrote()
  • Kernel SGD
    • kernel_sgd.HingeLoss()
    • kernel_sgd.KernelSGD()
    • kernel_sgd.GaussianKernel()

TODO

A number of packages have been identified but not been added yet. As a general rule for identifying potential projects to add to sklearn-extensions, if it cannot be pip-installed: it may be a candidate here

If you have any more suggestions, please feel free to add them, or let me know and I will try to.

Contributing

If you have an extension that you’d like to add, please submit a pull request and we can throw it in. A major benefit of this package is that we will aim to consolidate requirements among the disparate projects, therefore, for the sake of management, the code for the projects will be replicated here. In the spirit of OSS, we will also aim to contribute any meaningful changes back to the original projects as well.

A complete addition of a new package has a few components:

  • Actual addition of package into sklearn_extensions directory
  • Documentation of the included transformers/predictors in the sklearn_extensions docs
  • An example or two (included in the aforementioned docs as well) in the examples directory
  • A test or two, more if the source package has poor testing coverage

License

In most cases, all that sklearn_extensions does with external projects is include them. All of the projects will remain segregated into their own subdirectory, and will carry their original licenses in those subdirectories.

All material specific to this project (specifically any docs, tests, examples or original code) is released under the BSD 3-clause license. Any packages included in the bundle retain their original licences (as included in thir subdirectories)

Contents:

Indices and tables