Interpretable Machine Learning : A Guide for Making Black Box Models Explainable / Christoph MOLNAR

Titre : Interpretable Machine Learning : A Guide for Making Black Box Models Explainable Type de document : texte imprimé Auteurs : Christoph MOLNAR, Auteur Mention d'édition : 3ème éd. Editeur : Munich [Germany] : Christoph Molnar Année de publication : DL 2025 ISBN/ISSN/EAN : 978-3-911578-03-5 Note générale : Content About the book Appendices Catégories : Apprentissage automatique Arbres de décision Décomposition (méthode mathématique) Intelligence artificielle Réseaux neuronaux (informatique) Index. décimale : 006.3 Intelligence artificielle Résumé : Introduction Don't blindly trust model performance A young field with old roots How to read the book Interpretability Importance of interpretability Human-friendly explanations Goals of Interpretability Improving the model Justify model and predictions Discover insights Methods overview Interpretable models by design Post-hoc interpretability The lines are blurred Data and models Bike rentals (regression) Palmer penguins (classification) INTERPRETABLE MODELS Linear regression Interpretation Example Weight and effect plot Encoding categorical features Sparse linear models Strengths Limitations Logistic regression Don't use linear regression for classification Theory Interpretation Example Strengths Limitations Software GLM, GAM and more Non-Gaussian outcomes - GLMs Interactions Nonlinear effects - GAMs Strengths Limitations Software Further extensions Decision tree Interpretation Example Strengths Limitations Software Decision rules Learn rules from a single feature (OneR) Sequential covering Bayesian Rule Lists Strengths Limitations Software and alternatives Rulefit Interpretation and example Theory Strengths Limitations Software and alternatives LOCAL MODEL-AGNOSTIC METHOS Ceteris paribus plots Algorithm Examples Strengths Limitations Software and alternatives Individual conditional expectation (ice) Examples Centered ICE plot Derivative ICE plot Strengths Limitations Software and alternatives LIME LIME for tabular data LIME for text data LIME for image data Strengths Limitations Software Counterfactual explanations Generating counterfactual explanations Example Strengths Limitations Software and alternatives Scoped rules (Anchors) Finding anchors Complexity and runtime Tabular data example Strengths Limitations Software and alternatives Shapley values General idea Examples and interpretation Shapley value theory Estimating Shapley values Strengths Limitations Software and alternatives SHAP SHAP theory SHAP estimation Example SHAP aggregation plots Strengths Limitations Software GLOBAL MODEL-AGNOSTIC METHODS partial dependence plot (PDP) Definition and estimation Examples PDP-based feature importance Strengths Limitations Software and alternatives Accumulated local effects (ALE) Motivation and intuition Theory Estimation ALE versus PDP Examples Strengths Limitations Software and alternatives Feature interaction What are feature interactions? Friedman's H-statistic Examples Strengths Limitations Software and alternatives Functional decomposition Decomposing a function Functional ANOVA Generalized Functional ANOVA for dependent features Accumulated Local Effects Decomposing tree ensembles Statistical regression models Strengths Limitations Permutation feature importance Theory Example and interpretation Conditional feature importance Group-wise PFI example Strengths Limitations Software and alternatives Leave One Feature Out (LOFO) Importance Examples LOFO versus PFI Strengths Limitations Software and alternatives Surrogate models Theory Example Strengths Limitations Software Prototypes and criticisms Theory Examples Strengths Limitations Software and alternatives NEURAL NETWORK INTERPRETATION Learned features Feature visualization Network Dissection Strengths Limitations Software and further material Saliency maps Vanilla Gradient DeconvNet Grad-CAM Guided Grad-CAM SmoothGrad Examples Strengths Limitations Software Detecting concepts TCAV: Testing with Concept Activation Vectors Example Strengths Limitations Other concept-based approaches Software Adversarial examples Methods and examples The cybersecurity perspective Influential instances Deletion Diagnostics Influence Functions Strengths Limitations Software and Alternatives BEYOND THE METHODS Evaluation of interpretability methods Levels of evaluation Properties of explanations Story time The future of interpretability The future of machine learning The future of interpretability Translations Citing this book Acknowledgments

Exemplaires(1)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
B00012145	006.3 MOL	Ouvrage	BIBLIOTHÈQUE - ACCÈS LIBRE	000 - Informatique - Bibliothéconomie	Sorti jusqu'au 15/05/2028

---

1 (1 - 1 / 1) Par page : 25 50 100 200