Explainable and interpretable models in computer vision and machine learning

✍ Scribed by Baró, Xavier; Escalante, Hugo Jair; Escalera, Sergio; Gerven, Marcel van; Güçlü, Umut; Güçlütürk, Yağmur; Guyon, Isabelle et al. (eds.)

• Publisher: Springer
• Year: 2018
• Tongue: English
• Leaves: 305
• Series: Springer series on challenges in machine learning
• Category: Library

✦ Synopsis

This book compiles leading research on the development of explainable and interpretable machine learning methods in the context of computer vision and machine learning. Research progress in computer vision and pattern recognition has led to a variety of modeling techniques with almost human-like performance. Although these models have obtained astounding results, they are limited in their explainability and interpretability: what is the rationale behind the decision made? what in the model structure explains its functioning? Hence, while good performance is a critical required characteristic for learning machines, explainability and interpretability capabilities are needed to take learning machines to the next step to include them in decision support systems involving human supervision. This book, written by leading international researchers, addresses key topics of explainability and interpretability, including the following: · Evaluation and Generalization in Interpretable Machine Learning · Explanation Methods in Deep Learning · Learning Functional Causal Models with Generative Neural Networks · Learning Interpreatable Rules for Multi-Label Classification · Structuring Neural Networks for More Explainable Predictions · Generating Post Hoc Rationales of Deep Visual Classification Decisions · Ensembling Visual Explanations · Explainable Deep Driving by Visualizing Causal Attention · Interdisciplinary Perspective on Algorithmic Job Candidate Search · Multimodal Personality Trait Analysis for Explainable Modeling of Job Interview Decisions · Inherent Explainability Pattern Theory-based Video Event Interpretations.  Read more...

Abstract: This book compiles leading research on the development of explainable and interpretable machine learning methods in the context of computer vision and machine learning. Research progress in computer vision and pattern recognition has led to a variety of modeling techniques with almost human-like performance. Although these models have obtained astounding results, they are limited in their explainability and interpretability: what is the rationale behind the decision made? what in the model structure explains its functioning? Hence, while good performance is a critical required characteristic for learning machines, explainability and interpretability capabilities are needed to take learning machines to the next step to include them in decision support systems involving human supervision. This book, written by leading international researchers, addresses key topics of explainability and interpretability, including the following: · Evaluation and Generalization in Interpretable Machine Learning · Explanation Methods in Deep Learning · Learning Functional Causal Models with Generative Neural Networks · Learning Interpreatable Rules for Multi-Label Classification · Structuring Neural Networks for More Explainable Predictions · Generating Post Hoc Rationales of Deep Visual Classification Decisions · Ensembling Visual Explanations · Explainable Deep Driving by Visualizing Causal Attention · Interdisciplinary Perspective on Algorithmic Job Candidate Search · Multimodal Personality Trait Analysis for Explainable Modeling of Job Interview Decisions · Inherent Explainability Pattern Theory-based Video Event Interpretations

✦ Table of Contents

Content: Intro
Foreword
Preface
Acknowledgements
Contents
Contributors
Part I Notions and Concepts on Explainability and Interpretability
Considerations for Evaluation and Generalization in Interpretable Machine Learning
1 Introduction
2 Defining Interpretability
3 Defining the Interpretability Need
4 Evaluation
5 Considerations for Generalization
6 Conclusion: Recommendations for Researchers
References
Explanation Methods in Deep Learning: Users, Values, Concerns and Challenges
1 Introduction
1.1 The Components of Explainability
1.2 Users and Laws
1.3 Explanation and DNNs 2 Users and Their Concerns2.1 Case Study: Autonomous Driving
3 Laws and Regulations
4 Explanation
5 Explanation Methods
5.1 Desirable Properties of Explainers
5.2 A Taxonomy for Explanation Methods
5.2.1 Rule-Extraction Methods
5.2.2 Attribution Methods
5.2.3 Intrinsic Methods
6 Addressing General Concerns
7 Discussion
References
Part II Explainability and Interpretability in Machine Learning
Learning Functional Causal Models with Generative NeuralNetworks
1 Introduction
2 Problem Setting
2.1 Notations
2.2 Assumptions and Properties
3 State of the Art
3.1 Learning the CPDAG 3.1.1 Constraint-Based Methods3.1.2 Score-Based Methods
3.1.3 Hybrid Algorithms
3.2 Exploiting Asymmetry Between Cause and Effect
3.2.1 The Intuition
3.2.2 Restriction on the Class of Causal Mechanisms Considered
3.2.3 Pairwise Methods
3.3 Discussion
4 Causal Generative Neural Networks
4.1 Modeling Continuous FCMs with Generative Neural Networks
4.1.1 Generative Model and Interventions
4.2 Model Evaluation
4.2.1 Scoring Metric
4.2.2 Representational Power of CGNN
4.3 Model Optimization
4.3.1 Parametric (Weight) Optimization
4.3.2 Non-parametric (Structure) Optimization 4.3.3 Identifiability of CGNN up to Markov Equivalence Classes5 Experiments
5.1 Experimental Setting
5.2 Learning Bivariate Causal Structures
5.2.1 Benchmarks
5.2.2 Baseline Approaches
5.2.3 Hyper-Parameter Selection
5.2.4 Empirical Results
5.3 Identifying v-structures
5.4 Multivariate Causal Modeling Under Causal Sufficiency Assumption
5.4.1 Results on Artificial Graphs with Additive and Multiplicative Noises
5.4.2 Result on Biological Data
5.4.3 Results on Biological Real-World Data
6 Towards Predicting Confounding Effects
6.1 Principle
6.2 Experimental Validation 6.2.1 Benchmarks6.2.2 Baselines
6.2.3 Results
7 Discussion and Perspectives
Appendix
The Maximum Mean Discrepancy (MMD) Statistic
Proofs
Table of Scores for the Experiments on Cause-Effect Pairs
Table of Scores for the Experiments on Graphs
References
Learning Interpretable Rules for Multi-Label Classification
1 Introduction
2 Multi-Label Classification
2.1 Problem Definition
2.2 Dependencies in Multi-Label Classification
2.3 Evaluation of Multi-Label Predictions
2.3.1 Bipartition Evaluation Functions
2.3.2 Multi-Label Evaluation Functions
2.3.3 Aggregation and Averaging

✦ Subjects

Machine learning.;Computer vision.;Artificial Intelligence.;Image Processing and Computer Vision.;Pattern Recognition.