API Reference

This section contains the complete API reference for FDFI, auto-generated from the source code docstrings.

• Explainers
  • Overview
  • Shared Methods
  • Loss Functions
  • Base Explainer
  • Tree-Based Models
  • Linear Models
  • Kernel Methods
  • Gaussian Optimal Transport (OTExplainer)
  • Entropic Optimal Transport (EOTExplainer)
  • Shared Disentanglement Diagnostics
  • Flow-Based DFI (FlowExplainer)
  • DFIExplainer Alias
  • Cross-Fitting (Crossfitting)
  • Group Importance
• Plotting Functions
  • Overview
  • Global Importance
  • Summary Plot
  • Single-Explanation Views
  • Dependence Plot
  • Feature Correlation
  • Inference and Diagnostics
• Utilities
  • Overview
  • Statistical Utilities
  • Feature Type Detection
  • Gower Distance
  • Internal Helpers

Overview

FDFI provides the following main modules:

Explainers (fdfi.explainers)

Core explainer classes for computing feature importance. Includes base Explainer class, specialized explainers (TreeExplainer, LinearExplainer, KernelExplainer), optimal transport-based explainers (OTExplainer, EOTExplainer), flow-based DFI (FlowExplainer), and cross-fitted inference (Crossfitting).

Plotting (fdfi.plots)

Visualization functions for feature importance including summary plots, waterfall plots, force plots, and dependence plots.

Utilities (fdfi.utils)

Helper functions for input validation, data sampling, feature name management, and statistical utilities like the TwoComponentMixture class for variance floor estimation.