Interpretability for Industry 4.0 : statistical and machine learning approaches / / Antonio Lepore, Biagio Palumbo, Jean-Michel Poggi, editors

• Record Nr.: UNISA996495169103316
• Pubbl/distr/stampa: Cham, Switzerland : , : Springer, , [2022]; ©2022
• ISBN: 3-031-12402-2
• Descrizione fisica: 1 online resource (130 pages) : illustrations
• Disciplina: 658.4038028563
• Soggetti: Industry 4.0; Machine learning - Industrial applications; Industry 4.0 - Statistical methods; Aprenentatge automàtic; Aplicacions industrials; Llibres electrònics
• Lingua di pubblicazione: Inglese
• Formato: Materiale a stampa
• Livello bibliografico: Monografia
• Nota di bibliografia: Includes bibliographical references.
• Nota di contenuto: Intro -- Preface -- Contents -- 1 Different Views of Interpretability -- 1.1 Introduction -- 1.2 Interpretability: In Praise of Transparent Models -- 1.2.1 What Happened? -- 1.2.2 What Will Happen? -- 1.2.3 What Shall be Done to Make It Happen? -- 1.2.4 Patterns and Models -- 1.3 Generalizability and Interpretability with Industry 4.0 Implications -- 1.3.1 Introduction to Interpretable AI -- 1.3.2 A Wide Angle Perspective of Generalizability -- 1.3.3 Statistical Generalizability -- 1.4 Connections Between Interpretability in Machine Learning and Sensitivity Analysis of Model Outputs -- 1.4.1 Machine Learning and Uncertainty Quantification -- 1.4.2 Basics on Sensitivity Analysis and Its Main Settings -- 1.4.3 A Brief Taxonomy of Interpretability in Machine Learning -- 1.4.4 A Review of Sensitivity Analysis Powered Interpretability Methods -- References -- 2 Model Interpretability, Explainability and Trust for Manufacturing 4.0 -- 2.1 Manufacturing 4.0: Driving Trends for Data Mining -- 2.1.1 Process Monitoring in Manufacturing 4.0 -- 2.1.2 Design of Experiments in Manufacturing ; 4.0 -- 2.1.3 Increasing Trust in AI Models for Manufacturing 4.0: Interpretability, Explainability and Robustness -- 2.2 Additive Manufacturing as a Paradigmatic Example of Manufacturing 4.0 -- 2.3 Increase Trust in Additive Manufacturing: Robust Functional Analysis of Variance in Video-Image Analysis -- 2.3.1 The RoFANOVA Approach -- 2.3.2 An Additive Manufacturing Application -- References -- 3 Interpretability via Random Forests -- 3.1 Introduction -- 3.2 Interpretable Rule-Based Models -- 3.2.1 Literature Review -- 3.2.1.1 Definitions and Origins of Rule Models -- 3.2.1.2 Decision Trees -- 3.2.1.3 Tree-Based Rule Learning -- 3.2.1.4 Modern Rule Learning -- 3.2.2 SIRUS: Stable and Interpretable RUle Set -- 3.2.2.1 SIRUS Algorithm -- 3.2.2.2 Theoretical Analysis.; 3.2.2.3 Experiments -- 3.2.3 Discussion -- 3.3 Post-Processing of Black-Box Algorithms via Variable Importance -- 3.3.1 Literature Review -- 3.3.1.1 Model-Specific Variable Importance -- 3.3.1.2 Global Sensitivity Analysis -- 3.3.1.3 Local Interpretability -- 3.3.2 Sobol-MDA -- 3.3.2.1 Sobol-MDA Algorithm -- 3.3.2.2 Sobol-MDA Properties -- 3.3.2.3 Experiments -- 3.3.3 SHAFF: SHApley eFfects Estimates via Random Forests -- 3.3.3.1 SHAFF Algorithm -- 3.3.3.2 SHAFF Consistency -- 3.3.3.3 Experiments -- 3.3.4 Discussion -- References -- 4 Interpretability in Generalized Additive Models -- 4.1 GAMs: A Basic Framework for Flexible Interpretable Regression -- 4.1.1 Flexibility Can Be Important -- 4.1.2 Making the Model Computable -- 4.1.3 Estimation and Inference -- 4.1.4 Checking, Effective Degrees of Freedom and Model Selection -- 4.1.5 GAM Computation with mgcv in R -- 4.1.6 Smooths of Several Predictors -- 4.1.7 Further Interpretable Structure -- 4.2 From GAM to GAMLSS: Interpretability for Model Building -- 4.2.1 GAMLSS Modelling of UK Aggregate Electricity Demand -- 4.2.1.1 Data Overview and Pre-processing -- 4.2.1.2 Interactive GAMLSS Model Building -- 4.3 From GAMs to Aggregations of Experts, Are We Still Interpretable? -- 4.3.1 Online Forecasting with Online Aggregation of Experts -- 4.3.2 Visualizing the Black Boxes -- References.