Get started#
Install the library, run a first constrained model, and learn to read the guarantee it gives you.
pip install fair-seldonian
Python 3.10 or later. The core install pulls in NumPy, pandas, scikit-learn,
SciPy and PyTorch; [plots] adds matplotlib for the figures and
[notebook] adds JupyterLab for the example notebooks.
What it does and why
The problem Seldonian algorithms solve, what “high-confidence” means here, and why a refusal is a feature rather than a failure.
Install and run
Installation, a worked example end to end, reading QSAResult and its
diagnostics, and how to extend the framework to your own model.