altai 0.9.7.dev0

AffectLog's Trustworthy AI: Tools for model transparency, explainability, and regulatory compliance.

Table of Contents

• Installation and Usage
• Supported Methods
  • Model Explanations
  • Model Confidence
  • Prototypes
  • References and Examples
• Citations
• License and Notice

Installation and Usage

altai can be installed from:

• PyPI or GitHub source (using pip)
• Anaconda (using conda/mamba)

With pip

• To install altai from PyPI:

  pip install altai
  

• Alternatively, install the development version:

  pip install git+https://github.com/roy-saurabh/altai.git
  

• For distributed computation support (with ray):

  pip install altai[ray]
  

• For SHAP support:

  pip install altai[shap]
  

Usage

The altai explanation API is inspired by scikit-learn’s style, using distinct initialize, fit, and explain steps:

from affectlog_tai.explainers import AnchorTabular

# Initialize and fit the explainer with your prediction function and data
explainer = AnchorTabular(predict_fn, feature_names=feature_names, categorical_names=category_map)
explainer.fit(X_train)

# Explain an instance
explanation = explainer.explain(x)

The returned Explanation object contains meta (metadata and hyperparameters) and data (the computed explanation). For AnchorTabular, for example, you can access the anchor conditions via explanation.data['anchor'].

Supported Methods

Model Explanations

Method	Models	Explanations	Classification	Regression	Tabular	Text	Images	Categorical features	Train set required	Distributed
ALE	BB	global	✔	✔	✔
Partial Dependence	BB WB	global	✔	✔	✔			✔
PD Variance	BB WB	global	✔	✔	✔			✔
Permutation Importance	BB	global	✔	✔	✔			✔
Anchors	BB	local	✔		✔	✔	✔	✔	For Tabular
CEM	BB* TF/Keras	local	✔		✔		✔		Optional
Counterfactuals	BB* TF/Keras	local	✔		✔		✔		No
Prototype Counterfactuals	BB* TF/Keras	local	✔		✔		✔	✔	Optional
Counterfactuals with RL	BB	local	✔		✔		✔	✔	✔
Integrated Gradients	TF/Keras	local	✔	✔	✔	✔	✔	✔	Optional
Kernel SHAP	BB	local/global	✔	✔	✔			✔	✔	✔
Tree SHAP	WB	local/global	✔	✔	✔			✔	Optional
Similarity explanations	WB	local	✔	✔	✔	✔	✔	✔	✔

Model Confidence

These methods provide instance-specific scores that measure the model’s confidence.

Method	Models	Classification	Regression	Tabular	Text	Images	Categorical Features	Train set required
Trust Scores	BB	✔		✔	✔(1)	✔(2)		Yes
Linearity Measure	BB	✔	✔	✔		✔		Optional

Key:

• BB – Black-box (only require a prediction function)
• BB* – Black-box but assume model is differentiable
• WB – White-box (access to model internals)
• TF/Keras – TensorFlow models via the Keras API
• Local – Explains a single prediction
• Global – Explains overall model behavior
• (1) and (2) – Model-dependent requirements

Prototypes

These methods distill a dataset into a 1-KNN interpretable classifier.

Method	Classification	Regression	Tabular	Text	Images	Categorical Features	Train set labels
ProtoSelect	✔		✔	✔	✔	✔	Optional

---

References and Examples

• Accumulated Local Effects (ALE):

  • Documentation
  • Examples: California housing dataset, Iris dataset

• Partial Dependence:

  • Documentation
  • Example: Bike rental

• Permutation Importance:

  • Documentation
  • Example: Who's Going to Leave Next?

• Anchors:

  • Documentation
  • Examples: Adult income prediction, Iris dataset, Movie sentiment classification