Tutorials

This section contains hands-on tutorials for learning FDFI. Each tutorial is a Jupyter notebook that you can run interactively.

Getting Started

If you’re new to FDFI, start with the Quickstart tutorial to learn the basics in 5 minutes.

Quickstart: FDFI in 5 Minutes: Learn the basics of FDFI in 5 minutes. Create your first explainer, compute feature importance, and interpret the results.
OTExplainer: Gaussian Optimal Transport: Deep dive into the Gaussian OT explainer. Learn about the mathematical foundation, hyperparameters, shared diagnostics, and when to use it.
EOTExplainer: Semicontinuous Entropic Optimal Transport: Mixed-type-first EOT tutorial. Starts with active-feature screening using Gower cost, then covers epsilon, stochastic transport, target choice, and shared diagnostics.
FlowExplainer: Flow-Disentangled Feature Importance: Master Flow-DFI with normalizing flows. Learn about CPI vs SCPI methods, custom flow models, shared diagnostics, and when to choose FlowExplainer.
<no title>: Statistical inference with FDFI. Learn to compute confidence intervals, perform hypothesis testing, and identify significant features.

You can run these tutorials in several ways:

Option 1: Jupyter Notebook

cd docs/tutorials
jupyter notebook

Option 2: JupyterLab

cd docs/tutorials
jupyter lab

Option 3: VS Code

Open the .ipynb files directly in VS Code with the Jupyter extension.

Option 4: Google Colab

Upload the notebooks to Google Colab and run in the cloud.

Make sure you have FDFI installed with plotting support:

pip install -e ".[plots]"