Modeling Decisions for Artificial Intelligence: 19th International Conference, MDAI 2022, Sant Cugat, Spain, August 30 – September 2, 2022, Proceedings (Lecture Notes in Artificial Intelligence)

This book constitutes the refereed proceedings of the 19th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2022, held in Sant Cugat, Spain, during August - September 2022. The 16 papers presented in this volume were carefully reviewed and selected from 41 submissions. The papers discuss different facets of decision processes in a broad sense and present research in data science, machine learning, data privacy, aggregation functions, human decision-making, graphs and social networks, and recommendation and search. They were organized in topical sections as follows: Decision making and uncertainty; Data privacy; Machine Learning and data science. Preface 6 Organization 7 Invited Talks 10 Mathematical Modeling of COVID-19 from a Complex Systems Perspective 11 Explaining Black Box Classifiers by Exploiting Auto-Encoders 12 The Labor Impacts of Algorithmic Management 13 Contents 14 Decision Making and Uncertainty 16 Optimality Analysis for Stochastic LP Problems 17 1 Introduction 17 2 LP Problems and Robust Optimality Analysis 19 2.1 LP Problems with Uncetainties in Intervals 20 3 Stochastic Linear Programming 21 3.1 The Optimality Degree of an NBF Solution 22 3.2 Generation of Multiple NBF Solutions 24 3.3 Algorithm 24 4 Numerical Examples 25 5 Conclusions 27 References 27 A Multi-perceptual-Based Approach for Group Decision Aiding 29 1 Introduction 29 2 Hesitant Fuzzy Linguistic Term Sets for Decision Aiding 31 2.1 The Lattice of HFLTS 32 2.2 Concordance and Distance Between HFLTSs 32 3 The Perceptual Map on the Structure of the Lattice HS 33 3.1 Centroid and Degree of Consensus 34 4 A Transformation Function for Multi-perceptual GDM 35 5 Conclusion and Future Work 37 References 38 Probabilistic Judgement Aggregation by Opinion Update 40 1 Introduction 40 2 Framework 41 2.1 Probabilistic Judgement Profiles 41 2.2 Rationality of Probabilistic Judgement Sets 43 3 Updating Probabilistic Judgements 44 4 Convergence to a Consensus 46 5 A Necessary and Sufficient Condition for Reaching a Consensus 48 6 Conclusions 50 References 51 Semiring-Valued Fuzzy Rough Sets and Colour Segmentation 52 1 Introduction 52 2 Semiring-Valued Fuzzy Sets 53 3 Rough (R,R*)-Fuzzy Sets 56 4 Examples of Applications 59 4.1 (R2,R2*)-Fuzzy Rough Sets 59 4.2 (R1,R1*)-Fuzzy Rough Sets 60 4.3 Visualisation and Colour Segmentation 61 5 Conclusions 63 References 63 Data Privacy 65 Bistochastic Privacy 66 1 Introduction 66 2 Connections Between SDC, Differential Privacy and k-Anonymity Through Bistochastic Matrices 67 2.1 Connection with SDC 67 2.2 Connection with Differential Privacy 68 2.3 Connection with k-Anonymity 69 3 A Privacy Model Based on Bistochastic Matrices 71 3.1 Univariate Bistochastic Privacy 71 3.2 Bistochastic Privacy at the Data Set Level 73 4 Parameterization of Bistochastic Matrices 75 5 Conclusions and Future Research 77 Appendix 78 A.1 Randomized Response 78 A.2 The Permutation Model of SDC 78 A.3 Information Theory 79 References 80 Improvement of Estimate Distribution with Local Differential Privacy 81 1 Introduction 81 2 Local Differential Privacy 83 2.1 Fundamental Definition 83 2.2 RAPPOR ch6RAPPOR 83 2.3 Related Works 84 3 Improvement of Estimate 85 3.1 Maximum Likelihood Estimate 85 3.2 Iterative Estimate 85 4 Experiment 87 4.1 Objective 87 4.2 Data 87 4.3 Method 88 4.4 Results 89 5 Conclusion 91 References 92 Geolocated Data Generation and Protection Using Generative Adversarial Networks 93 1 Introduction 94 2 Related Works 94 3 Background 96 3.1 GAN and W-GAN 96 3.2 Differential Privacy 97 3.3 DP-GAN 98 4 Methodology 98 5 Experiment and Results 99 6 Conclusions and Future Works 103 References 103 Machine Learning and Data Science 105 A Strategic Approach Based on AND-OR Recommendation Trees for Updating Obsolete Information 106 1 Introduction 106 2 Formal Background 108 2.1 Notation 108 2.2 Obsolete Information Detection System 109 3 Recommender System 110 3.1 Recommendations on How to Update Obsolete Information 110 3.2 Predictions 113 4 Empirical Results 114 4.1 Experimental Data 114 4.2 Results and Discussion 115 5 Conclusion 117 References 117 Identification of Subjects Wearing a Surgical Mask from Their Speech by Means of X-vectors and Fisher Vectors 119 1 Introduction 119 2 Data 121 3 Feature Extraction and Evaluation Methods 121 3.1 Frame-Level Features 121 3.2 X-vectors 122 3.3 Fisher Vectors 123 3.4 Support Vector Machines (SVM) 125 4 Experimental Setup 125 5 Results and Discussion 126 6 Conclusions 127 References 127 Measuring Fairness in Machine Learning Models via Counterfactual Examples 130 1 Introduction 130 2 Related Work 131 2.1 Fairness 131 2.2 Adversarial Examples 132 2.3 GANs 132 3 Measuring Fairness via Counterfactual Examples 133 3.1 Measuring Fairness in Tabular Data 133 3.2 Measuring Fairness in Image Data 134 4 Empirical Results 135 4.1 Experimental Setup 135 4.2 Results 138 5 Conclusion and Future Work 141 References 141 Re-calibrating Machine Learning Models Using Confidence Interval Bounds 143 1 Introduction 143 2 Method 144 2.1 Experimental Analysis 147 3 Results 149 4 Discussion 150 5 Conclusion 151 References 152 An Analysis of Byzantine-Tolerant Aggregation Mechanisms on Model Poisoning in Federated Learning 154 1 Introduction 154 2 Background 155 2.1 Federated Learning 155 2.2 Byzantine-Tolerant Aggregation 155 2.3 Related Work 156 3 Threat Model 156 3.1 Backdoor Attacks 157 3.2 Model Replacement 157 4 Experiments 157 4.1 Experimental Setup 157 4.2 Datasets and Learning Models 158 4.3 Experimental Fairness 159 5 Experimental Results 159 5.1 Aggregation Mechanisms and Defenses 160 5.2 Defense Fairness 163 6 Discussion and Conclusion 164 References 165 Effective Early Stopping of Point Cloud Neural Networks 167 1 Introduction 167 2 Related Work 168 3 Early Stopping of Point Cloud Neural Networks 170 3.1 Early Stopping Criteria 170 3.2 The Selection of the Stop-Window 171 4 Evaluation 172 4.1 Evaluation Protocol 173 4.2 Data and Models 173 4.3 Analysis of Our Proposal 173 4.4 Comparison to Conventional Early Stopping Techniques 174 5 Conclusion 177 References 177 Representation and Interpretability of IE Integral Neural Networks 179 1 Introduction 179 2 Preliminaries 180 3 Inclusion-Exclusion Integral Network 181 3.1 Network Implementation 182 3.2 Additive Representation of the Interaction Operator 183 3.3 Other Network Models 184 4 Application to Regression and Classification Datasets 184 4.1 Dataset: Boston House-Price Data 184 4.2 Experiment 1 – Comparison of Two Equivalent Networks 184 4.3 Experiment 2 – Comparison of Interaction Operators 186 5 Discussion 189 References 190 Deep Attributed Graph Embeddings 192 1 Introduction 192 2 Related Works 193 3 Deep Attributed Graph Embedding (DAGE) 195 3.1 Basic Definitions and Preliminaries 195 3.2 The Proposed Model 195 4 Experimental Analysis 198 4.1 Node Classification 199 4.2 Node Clustering 199 4.3 Trainable Parameters 201 5 Conclusions and Future Work 201 References 202 Estimation of Prediction Error with Regression Trees 204 1 Introduction 204 2 Regression Trees 206 3 Methodology 207 3.1 Example 208 4 Experiments 209 5 Conclusions 212 References 213 Author Index 214 This book constitutes the refereed proceedings of the 9th International Conference on Modeling Decisions for Artificial Intelligence, MDAI 2012, held in Girona, Catalonia, Spain, in November 2012. The 32 revised full papers were carefully reviewed and selected from 49 submissions and are presented with 4 plenary talks. The papers are organized in topical sections on aggregation operators, integrals, data privacy and security, reasoning, applications, and clustering and similarity.