This repository is fully operational on Ubuntu 19.10. On MacOSX only CPU functionality has been tested.

pip install prob_forest

conda env create --file env.yml
conda activate prob_forest

This package was developed using torch 1.3.1 and numpy 1.17.4 under python 3.7.5.

This package uses the latest development version of sklearn, which can be installed from source here.

pytest is required to run tests.

pandas and tqdm required for certain scripts.

pip install -e .

make test

There are two considerations, which are motivated by keeping to the sklearn API:

1. All parameter names are given in the singular (ex. 'split' instead of 'splits')
2. Snake case, with separate prefix (ex. n_feature instead of nfeature)