180 lines (171 loc) · 11.9 KB

Table of contents

Overview

Explanations

Methods
Examples
- Alibi Overview Examples
- Accumulated Local Effets
  - Accumulated Local Effects for classifying flowers
  - Accumulated Local Effects for predicting house prices
- Anchors
- Contrastive Explanation Method
  - Contrastive Explanations Method (CEM) applied to Iris dataset
  - Contrastive Explanations Method (CEM) applied to MNIST
- Counterfactual Instances on MNIST
- Counterfactuals Guided by Prototypes
- Counterfactuals with Reinforcement Learning
  - Counterfactual with Reinforcement Learning (CFRL) on Adult Census
  - Counterfactual with Reinforcement Learning (CFRL) on MNIST
- Integrated Gradients
- Kernel SHAP
- Partial Dependence
  - Partial Dependence and Individual Conditional Expectation for predicting bike renting
- Partial Dependence Variance
  - Feature importance and feature interaction based on partial dependece variance
- Permutation Importance
  - Permutation Feature Importance on “Who’s Going to Leave Next?”
- Similarity explanations
- Tree SHAP
  - Explaining Tree Models with Interventional Feature Perturbation Tree SHAP
  - Explaining Tree Models with Path-Dependent Feature Perturbation Tree SHAP

Model Confidence

Methods
- Measuring the linearity of machine learning models
- Trust Scores
Examples
- Measuring the linearity of machine learning models
  - Linearity measure applied to fashion MNIST
  - Linearity measure applied to Iris
- Trust Scores
  - Trust Scores applied to Iris
  - Trust Scores applied to MNIST

Prototypes

Methods
- ProtoSelect
Examples
- ProtoSelect on Adult Census and CIFAR10

API Reference