XAI Bugs and Experiments

This repository contains Jupyter notebooks for experiments and bug investigations related to Explainable AI (XAI) libraries, including SHAP, InterpretML, LightGBM, and FastTreeSHAP.

Notebooks Overview

• BayesianOptimization.ipynb
  Experiments with Bayesian optimization techniques.

• lgbm_objective.ipynb
  Investigates SHAP explanations for LightGBM models, including objective-related issues.

• lgbm_objective_fast.ipynb
  Uses FastTreeSHAP for efficient SHAP value computation on tree models.

• SuccessfulAdditiveExplainer.ipynb
  Demonstrates a successful use of SHAP's Additive explainer with InterpretML's ExplainableBoostingClassifier.

• UnsuccessfulAdditiveExplainer.ipynb
  Documents a failure case when using SHAP's Additive explainer with InterpretML models.

License

MIT License