دستورالعمل‌های هوش مصنوعی قابل توضیح: پیاده‌سازی راه‌حل‌هایی برای توضیح‌پذیری و تفسیرپذیری مدل با... پایتون

Understand how to use Explainable AI (XAI) libraries and build trust in AI and machine learning models. This book utilizes a problem-solution approach to explaining machine learning models and their algorithms. The book starts with model interpretation for supervised learning linear models, which includes feature importance, partial dependency analysis, and influential data point analysis for both classification and regression models. Next, it explains supervised learning using non-linear models and state-of-the-art frameworks such as SHAP values/scores and LIME for local interpretation. Explainability for time series models is covered using LIME and SHAP, as are natural language processing-related tasks such as text classification, and sentiment analysis with ELI5, and ALIBI. The book concludes with complex model classification and regression-like neural networks and deep learning models using the CAPTUM framework that shows feature attribution, neuron attribution, and activation attribution. After reading this book, you will understand AI and machine learning models and be able to put that knowledge into practice to bring more accuracy and transparency to your analyses. What You Will Learn Create code snippets and explain machine learning models using Python Leverage deep learning models using the latest code with agile implementations Build, train, and explain neural network models designed to scale Understand the different variants of neural network models Who This Book Is For AI engineers, data scientists, and software developers interested in XAI Table of Contents 4 About the Author 16 About the Technical Reviewer 17 Acknowledgments 18 Introduction 19 Chapter 1: Introducing Explainability and Setting Up Your Development Environment 21 Recipe 1-1. SHAP Installation 23 Problem 23 Solution 23 How It Works 24 Recipe 1-2. LIME Installation 26 Problem 26 Solution 26 How It Works 26 Recipe 1-3. SHAPASH Installation 28 Problem 28 Solution 28 How It Works 29 Recipe 1-4. ELI5 Installation 29 Problem 29 Solution 29 How It Works 29 Recipe 1-5. Skater Installation 31 Problem 31 Solution 31 How It Works 31 Recipe 1-6. Skope-rules Installation 32 Problem 32 Solution 32 How It Works 32 Recipe 1-7. Methods of Model Explainability 33 Problem 33 Solution 33 How It Works 34 Conclusion 35 Chapter 2: Explainability for Linear Supervised Models 36 Recipe 2-1. SHAP Values for a Regression Model on All Numerical Input Variables 37 Problem 37 Solution 37 How It Works 37 Recipe 2-2. SHAP Partial Dependency Plot for a Regression Model 44 Problem 44 Solution 44 How It Works 44 Recipe 2-3. SHAP Feature Importance for Regression Model with All Numerical Input Variables 48 Problem 48 Solution 48 How It Works 48 Recipe 2-4. SHAP Values for a Regression Model on All Mixed Input Variables 50 Problem 50 Solution 51 How It Works 51 Recipe 2-5. SHAP Partial Dependency Plot for Regression Model for Mixed Input 54 Problem 54 Solution 55 How It Works 55 Recipe 2-6. SHAP Feature Importance for a Regression Model with All Mixed Input Variables 60 Problem 60 Solution 60 How It Works 60 Recipe 2-7. SHAP Strength for Mixed Features on the Predicted Output for Regression Models 62 Problem 62 Solution 62 How It Works 62 Recipe 2-8. SHAP Values for a Regression Model on Scaled Data 63 Problem 63 Solution 63 How It Works 64 Recipe 2-9. LIME Explainer for Tabular Data 67 Problem 67 Solution 68 How It Works 68 Recipe 2-10. ELI5 Explainer for Tabular Data 70 Problem 70 Solution 70 How It Works 70 Recipe 2-11. How the Permutation Model in ELI5 Works 72 Problem 72 Solution 72 How It Works 73 Recipe 2-12. Global Explanation for Logistic Regression Models 73 Problem 73 Solution 73 How It Works 74 Recipe 2-13. Partial Dependency Plot for a Classifier 77 Problem 77 Solution 77 How It Works 77 Recipe 2-14. Global Feature Importance from the Classifier 80 Problem 80 Solution 80 How It Works 80 Recipe 2-15. Local Explanations Using LIME 82 Problem 82 Solution 82 How It Works 82 Recipe 2-16. Model Explanations Using ELI5 86 Problem 86 Solution 86 How It Works 86 Conclusion 90 References 91 Chapter 3: Explainability for Nonlinear Supervised Models 92 Recipe 3-1. SHAP Values for Tree Models on All Numerical Input Variables 93 Problem 93 Solution 93 How It Works 93 Recipe 3-2. Partial Dependency Plot for Tree Regression Model 100 Problem 100 Solution 100 How It Works 100 Recipe 3-3. SHAP Feature Importance for Regression Models with All Numerical Input Variables 101 Problem 101 Solution 102 How It Works 102 Recipe 3-4. SHAP Values for Tree Regression Models with All Mixed Input Variables 104 Problem 104 Solution 104 How It Works 104 Recipe 3-5. SHAP Partial Dependency Plot for Regression Models with Mixed Input 106 Problem 106 Solution 106 How It Works 107 Recipe 3-6. SHAP Feature Importance for Tree Regression Models with All Mixed Input Variables 109 Problem 109 Solution 110 How It Works 110 Recipe 3-7. LIME Explainer for Tabular Data 112 Problem 112 Solution 112 How It Works 112 Recipe 3-8. ELI5 Explainer for Tabular Data 115 Problem 115 Solution 115 How It Works 115 Recipe 3-9. How the Permutation Model in ELI5 Works 119 Problem 119 Solution 120 How It Works 120 Recipe 3-10. Global Explanation for Decision Tree Models 120 Problem 120 Solution 120 How It Works 121 Recipe 3-11. Partial Dependency Plot for a Nonlinear Classifier 123 Problem 123 Solution 123 How It Works 123 Recipe 3-12. Global Feature Importance from the Nonlinear Classifier 126 Problem 126 Solution 126 How It Works 126 Recipe 3-13. Local Explanations Using LIME 127 Problem 127 Solution 128 How It Works 128 Recipe 3-14. Model Explanations Using ELI5 132 Problem 132 Solution 132 How It Works 133 Conclusion 136 Chapter 4: Explainability for Ensemble Supervised Models 137 Recipe 4-1. Explainable Boosting Machine Interpretation 138 Problem 138 Solution 138 How It Works 139 Recipe 4-2. Partial Dependency Plot for Tree Regression Models 143 Problem 143 Solution 143 How It Works 143 Recipe 4-3. Explain a Extreme Gradient Boosting Model with All Numerical Input Variables 149 Problem 149 Solution 149 How It Works 149 Recipe 4-4. Explain a Random Forest Regressor with Global and Local Interpretations 154 Problem 154 Solution 154 How It Works 154 Recipe 4-5. Explain the Catboost Regressor with Global and Local Interpretations 157 Problem 157 Solution 157 How It Works 158 Recipe 4-6. Explain the EBM Classifier with Global and Local Interpretations 160 Problem 160 Solution 160 How It Works 161 Recipe 4-7. SHAP Partial Dependency Plot for Regression Models with Mixed Input 163 Problem 163 Solution 163 How It Works 163 Recipe 4-8. SHAP Feature Importance for Tree Regression Models with Mixed Input Variables 167 Problem 167 Solution 167 How It Works 168 Recipe 4-9. Explaining the XGBoost Model 172 Problem 172 Solution 172 How It Works 172 Recipe 4-10. Random Forest Regressor for Mixed Data Types 177 Problem 177 Solution 177 How It Works 177 Recipe 4-11. Explaining the Catboost Model 180 Problem 180 Solution 180 How It Works 180 Recipe 4-12. LIME Explainer for the Catboost Model and Tabular Data 183 Problem 183 Solution 183 How It Works 184 Recipe 4-13. ELI5 Explainer for Tabular Data 186 Problem 186 Solution 186 How It Works 186 Recipe 4-14. How the Permutation Model in ELI5 Works 190 Problem 190 Solution 190 How It Works 190 Recipe 4-15. Global Explanation for Ensemble Classification Models 191 Problem 191 Solution 191 How It Works 191 Recipe 4-16. Partial Dependency Plot for a Nonlinear Classifier 194 Problem 194 Solution 194 How It Works 194 Recipe 4-17. Global Feature Importance from the Nonlinear Classifier 196 Problem 196 Solution 196 How It Works 196 Recipe 4-18. XGBoost Model Explanation 199 Problem 199 Solution 199 How It Works 199 Recipe 4-19. Explain a Random Forest Classifier 207 Problem 207 Solution 207 How It Works 208 Recipe 4-20. Catboost Model Interpretation for Classification Scenario 210 Problem 210 Solution 211 How It Works 211 Recipe 4-21. Local Explanations Using LIME 212 Problem 212 Solution 213 How It Works 213 Recipe 4-22. Model Explanations Using ELI5 216 Problem 216 Solution 216 How It Works 216 Recipe 4-23. Multiclass Classification Model Explanation 219 Problem 219 Solution 219 How It Works 220 Conclusion 223 Chapter 5: Explainability for Natural Language Processing 225 Recipe 5-1. Explain Sentiment Analysis Text Classification Using SHAP 226 Problem 226 Solution 226 How It Works 226 Recipe 5-2. Explain Sentiment Analysis Text Classification Using ELI5 231 Problem 231 Solution 231 How It Works 231 Recipe 5-3. Local Explanation Using ELI5 234 Problem 234 Solution 234 How It Works 234 Conclusion 236 Chapter 6: Explainability for Time-Series Models 237 Recipe 6-1. Explain Time-Series Models Using LIME 238 Problem 238 Solution 238 How It Works 238 Recipe 6-2. Explain Time-Series Models Using SHAP 246 Problem 246 Solution 246 How It Works 246 Conclusion 249 Chapter 7: Explainability for Deep Learning Models 250 Recipe 7-1. Explain MNIST Images Using a Gradient Explainer Based on Keras 251 Problem 251 Solution 251 How It Works 251 Recipe 7-2. Use Kernel Explainer–Based SHAP Values from a Keras Model 256 Problem 256 Solution 256 How It Works 257 Recipe 7-3. Explain a PyTorch-Based Deep Learning Model 260 Problem 260 Solution 260 How It Works 260 Conclusion 266