Welcome to REP’s documentation!

REP (Reproducible Experiment Platform) library provides functionality for all basic needs to deal with machine learning.

It includes:

  • Estimators (classification and regression) is sklearn-like wrappers for variety of machine learning libraries:

    • TMVA
    • Sklearn
    • XGBoost
    • Pybrain
    • Neurolab
    • Theanets.
    • MatrixNet service (available to CERN)

    These can be used as base estimators in sklearn.

  • Meta Machine Learning contains factory (the set of estimators), grid search, folding algorithm. Also parallel execution on a cluster is supported.

  • Metrics implement some basis for metrics used in reports and during grid search.

  • Report for models contains helpful classes to get model result information on any dataset.

  • Plotting is a wrapper for different plotting libraries including interactive plots.

    • matplotlib
    • bokeh
    • tmva
  • Parallel computing describes REP way to parallelize tasks.

  • Data defines LabeledDataStorage - a custom way to store training data in a single object.

  • Utilities contains additional functions.

  • REProducibility is a recipe to make research reliable.

  • Howto notebooks contains examples.

Main repository: http://github.com/yandex/rep


REP provides several installation ways.

Please find instructions at the repository main page.