Kowalski

A prosonal collection of routines.

• Includes custom collection of metrics and scorers (namely kappa and specificity on kowalski.utils.metrics)
• A collection of estimators with hyperparam grids to cross validate
• Soon to be included, a collection of feature selection methods with grids to be cross validated (kowalski.grid_search.estimator_pipelines.estimators_collection)
• somewhat mlflow oriented
• plotting wrappers, such as kowalski.utils.plotting.plot_confusion_matrix (much wow here)
• scoring convenience provided by kowalski.utils.evaluate.evaluate_on_test
• assessment interfaces that I haven't seen anywhere else - check the assessments section

installation instructions

Git clone this repo and pip install in dev mode to your target virtualenv/condaenv. Example:

git clone [email protected]:bgalvao/kowalski.git
cd kowalski
pip install -e ./  # fyi, this is run in the directory where setup.py resides

example usage

There are two main features of this pseudo-package.

• Nested Cross Validation assessment (kowalski.assessment.nested_cv.NestedCVTestAssessment)
• Target Shuffiling assessment (kowalski.assessment.target_shuffling.TargetShufflingAssessment)

(ok I'll write this later)