/cortado

High performance ML in pure Python

Primary LanguagePythonMIT LicenseMIT

cortado: high performance 100% Python package for machine learning

Installation

cortado can be installed from pip:

pip install cortado

Main features:

  • native support for both numeric and categorical data (covariates and factors)
  • innovative feature engineering: virtual data columns and easy conversions between numeric and categorical data
  • out of core data processing when dataframes are bigger than RAM
  • implementation of XGBoost logistic in 500 lines of Python code, 3x faster than original C++ implementation (using Numba jit under the hood)
  • easy to extend, written in functional style for easy composition
  • works well with pandas dataframes
  • more to come soon!

Demo notebooks on Kaggle:

How to contribute

All contributions and bug reports are welcome.

I'm available for hire!