Skip to main content

​
New features

FPDE is on PyPI. The first public release (v0.1.0) of the fpde package is available. Install it with python -m pip install fpde. See the quickstart to run your first explanation. Prototype-contrast explanations. FPDE builds class-mean prototypes from training data and explains target-versus-rival evidence with Diff-FPDE, Cos-FPDE, and Hyb-FPDE. Validation helpers. Use FPDEEngine.select_lambda to choose lambda_hyb on held-out validation data. Use perturbation_curves to compute deletion and insertion curves for an attribution vector.

​
Updates

Dual license. The package is distributed under MIT OR Apache-2.0. You may choose either license. Continuous publishing. Releases flow through automated publishing to PyPI.