Skip to main content
Ctrl+K

UnifiedIG 0.1.6

  • GitHub
  • GitHub

Documentation

User guide

  • Getting started
  • Baselines and CBaseline
  • Reading an explanation
  • Classification: attribute scores, not probabilities
  • Plot with SHAP
  • Supported models and backend selection
  • Choosing the attribution feature space
  • Framework adapters
  • Worked examples
  • Accuracy and troubleshooting
  • Loss attribution

Concepts

  • How UnifiedIG works
  • UnifiedIG semantics
  • The Integrated Gradients Stack

Reference and development

  • API reference
  • References and citation
  • Release notes
  • Roadmap
  • Building the documentation
  • Publishing releases
  • Public release checklist

Index

_ | C | E | J | L | M | N | T

_

  • __call__() (unifiedig.Explainer method)
    • (unifiedig.LossExplainer method)

C

  • contrast() (unifiedig.Explanation method)

E

  • Explainer (class in unifiedig)
  • Explanation (class in unifiedig)

J

  • JaxModel (class in unifiedig)

L

  • LossExplainer (class in unifiedig)

M

  • max_abs_completeness_error (unifiedig.Explanation property)

N

  • n_steps (unifiedig.LossExplainer property)

T

  • TensorFlowModel (class in unifiedig)
  • to_shap() (unifiedig.Explanation method)

© Copyright 2026, Ludger Hentschel.

Created using Sphinx 8.2.3.

Built with the PyData Sphinx Theme 0.16.1.