Learning and Intelligent Optimization : 16th International Conference, LION 16, Milos Island, Greece, June 5–10, 2022, Revised Selected Papers

This book constitutes the refereed proceedings of the 16th International Conference on Learning and Intelligent Optimization, LION 16, which took place in Milos Island, Greece, in June 2022. The 36 full papers and 3 short papers presented in this volume were carefully reviewed and selected from 60 submissions. LION deals with automatic solver configuration, parallel methods, intelligent optimization, nature-inspired algorithms, hard combinatorial optimization problems, DC learning, computational intelligence, and others. The contributions were organized in topical sections as follows: Invited Papers; Contributed Papers. Preface Organization Contents Invited Papers Optimal Scheduling of the Leaves of a Tree and the SVO Frequencies of Languages 1 Introduction: The Leaf Scheduling Problem 2 The Greedy Algorithm 3 Generating Sentences in the Brain 3.1 Scheduling Cost Explains SVO Frequencies 3.2 Leaf Scheduling Cost as Energy 3.3 A Statistical-Mechanical Argument 4 Discussion References From Design of Experiments to Combinatorics of Disasters: A Conceptual Framework for Disaster Exercises 1 Introduction 2 Motivation 2.1 Design of Experiments 2.2 Exercise Scenarios 2.3 Event Coverage 3 A Conceptual Combinatorial Framework for Disaster Scenario Generation 3.1 Exercise Formulation 3.2 Acquisition of Data 3.3 Exercise Design 3.4 Exercise Generation 3.5 Execution and Assessment 3.6 Post-processing 4 Conclusion and Future Work References Separating Two Polyhedra Utilizing Alternative Theorems and Penalty Function 1 Introduction 2 Separation of Two Polyhedra 3 Algorithm 4 Numerical Results 5 Conclusion References Contributed Papers A Composite Index Method for Optimization Benchmarking 1 Introduction 2 Performance Evaluation Using Composite Index Method (CIM) 3 The Backhaul Profit Maximization Problem (BPMP) 3.1 Node-arc Formulation 3.2 Triples Formulation 4 Node-Arc Summary 4.1 Computing Environment and Weight Selection 4.2 Initial Incumbent Formulation 4.3 Technique 1: Conditional Arc Flow 4.4 Technique 2: Relax Node-Degree Constraints 4.5 Technique 3: Single-Node Demand Cuts 4.6 Best Node-Arc Model 5 Triples Summary 6 Conclusions References Optimal Energy Management of Microgrid Using Multi-objective Optimisation Approach 1 Introduction 2 Microgrid Description 3 Modeling of the Energy Storage System 4 Optimisation Problem 4.1 Objective Functions 4.2 Constraints Functions 4.3 The Multi-objective Optimisation Problem 5 Pareto-Search Algorithm 6 Numerical Results and Discussion 6.1 Economic Criterion 6.2 Environmental Criterion 7 Conclusions and Future Work References A Stochastic Alternating Balance k-Means Algorithm for Fair Clustering 1 Introduction 2 The Mini-Batch k-means Algorithm 3 A New Stochastic Alternating Balance Fair k-means Method 3.1 The Bi-objective Balance k-means Formulation 3.2 The Stochastic Alternating Balance Fair k-means Method 4 Numerical Experiments 4.1 Pareto Front SAfairKM Algorithm 4.2 Numerical Results 5 Concluding Remarks A Description of an Existing Approach for Comparison B More Numerical Results References Binary Black Widow Optimization Algorithm for Feature Selection Problems 1 Introduction 2 BBWO Algorithm for FS 2.1 The Pseudo-code of the Proposed Method (BBWO) 3 Experiments 3.1 Implementation Setup 3.2 Evaluation Criteria and Parameters Setting 3.3 Experiment Results and Discussion of BBWO 4 Conclusion and Future Works References Learning to Solve a Stochastic Orienteering Problem with Time Windows 1 Introduction 2 Related Work 3 Background 4 The TDOP 5 Solving the TDOP 5.1 POMO for the TDOP 5.2 EAS for Stochastic Problems 5.3 Solution Construction Using Monte Carlo Rollouts 6 Computational Results 6.1 RQ1: Test Set Performance 6.2 RQ2: EAS 6.3 RQ3: Monte Carlo Rollouts 7 Conclusion References ML-Based Approach for Accelerating Global Search Algorithm for Solving Multicriteria Problems 1 Introduction 2 Problem Statement 3 General Computational Scheme 4 Approaches to Improving Search Efficiency 5 Results of Computational Experiments References The Skewed Kruskal Algorithm 1 Kruskal Algorithm 2 Improvements 2.1 QuickKruskal Algorithm 2.2 FilterKruskal Algorithm 2.3 SkewedFilterKruskal Algorithm 3 Computational Results 4 Conclusions and Open Questions References Bounds for Sparse Solutions of K-SVCR Multi-class Classification Model 1 Introduction 2 K-support Vector Classification Regression 3 Lower Bound for Nonzero Components of Solutions 4 Conclusion References Integer Linear Programming in Solving an Optimization Problem at the Mixing Department of the Metallurgical Production 1 Introduction 2 Statement of the Problem 3 Variable Set 4 Coefficients 5 Constraints 6 Objective 7 Computational Results 8 Conclusion References Realtime Gray-Box Algorithm Configuration 1 Introduction 2 Automated AC 2.1 Offline AC 2.2 RAC 3 Gray-box Method 3.1 Identifying Underperforming Configurations 3.2 Applying Cost-Sensitive Classification 3.3 Terminating Underperforming Configurations 3.4 Utilizing Freed Resources 4 Computational Experiments 4.1 Dataset and Solver 4.2 RQ1: Does CPPL Find the Best Configuration? 4.3 RQ2: Quality of Prediction Based on Gray-Box Data 4.4 RQ3: Black-Box vs. Gray-Box CPPL 5 Conclusion and Future Work References Dynamic Urban Solid Waste Management System for Smart Cities 1 Introduction 2 Related Literature 2.1 Waste Collection in Smart Cities 2.2 Waste Collection Route Problems 2.3 Algorithms for Route Optimization 3 Methodology 3.1 Problem Assembly 3.2 Level of Dumpsters 3.3 Dynamic Selection 4 Results and Discussion 4.1 Waste Level Throughout Days 4.2 Numerical Results Discussion 5 Conclusions and Future Work References Single MCMC Chain Parallelisation on Decision Trees 1 Introduction 2 Markov Chain Monte Carlo in General and Most Recent Work 2.1 Probabilistic Trees Packages and Level of Parallelism 3 Markov Chain Monte Carlo in Decision Tree 3.1 Specification of the Metropolis-Hastings Search Algorithm on Decision Trees 4 Parallelising a Single Decision Tree MCMC Chain 5 Results 5.1 Quality of the Samples Between Serial and Parallel Implementation 5.2 Practical Gains 6 Conclusion References An Extension of NSGA-II for Scaling up Multi-objective Spatial Zoning Optimization 1 Introduction 2 Related Work 3 Problem Resolution 3.1 Non-dominated Sorting Genetic Algorithm-II (NSGA-II) 3.2 Solution Encoding Schema 3.3 The Initialization Operators 3.4 Crossover Operators 3.5 Mutation Operators 3.6 Check and Repair Operators 3.7 Evaluation and Selection Operators 3.8 Stop Condition 4 Response Surface Methodology for Parameters Tuning 4.1 Multi-Response RSM (MRRSM) Optimization 4.2 Final Tuned Parameters 5 Results 5.1 Performance Measures 5.2 Comparison Analysis 6 Conclusion References Competitive Supply Allocation in a Distribution Network Under Overproduction 1 Introduction 2 Equilibrium Flow Allocation in a Single-Commodity Network 3 Competitive Supply Allocation in a Distribution Network Under Overproduction 4 Strategies of Suppliers Under Overproduction 5 Conclusion References Safe-Exploration of Control Policies from Safe-Experience via Gaussian Processes 1 Introduction 1.1 Motivation 1.2 Related Works 1.3 Contribution 2 Reference Problem 2.1 Control of a Dynamic System 2.2 Problem Formulation and Usual Solving Methods 3 Novel Approach: Safe-Exploring from Safe Experience 4 Case Studies 4.1 Case Study 1: Optimal Control of a Water Tank 4.2 Case Study 2: Optimal Control of a House Heating System 5 Experiments and Results 5.1 Results on Case Study 1 5.2 Results on Case Study 2 5.3 Software and Data 6 Conclusions, Limitations, and Perspectives References Bayesian Optimization in Wasserstein Spaces 1 Introduction 1.1 Related Works 1.2 Our Contributions 1.3 Organization of the Paper 2 Background 2.1 Wasserstein Distance 2.2 Bayesian Optimization 3 The WST-BO Algorithm 3.1 BO in the Wasserstein Space 3.2 Mapping W into X 4 Computational Results 4.1 Convergence 5 Conclusions and Perspectives References Network Vulnerability Analysis in Wasserstein Spaces 1 Introduction 1.1 Related Works 1.2 Our Contributions 2 Wasserstein 2.1 Basic Definitions 2.2 The Space of Quantile Functions 2.3 The Wasserstein Distance for Discrete Distributions: The Optimal Transport Approach 2.4 The Wasserstein Distance for Discrete Distributions: A Statistical Approach 2.5 Barycenter and Clustering 3 Distributional Representation of Networks 4 Vulnerability Measures 4.1 Efficiency-Based Vulnerability 4.2 Wasserstein-Based Vulnerability 5 Data and Software Resources 5.1 Networks 5.2 Software 6 Computational Results 6.1 Vulnerability 6.2 Clustering 7 Conclusions and Perspectives References BERT Self-Learning Approach with Limited Labels for Document Classification 1 Introduction 2 Related Work 3 Methodology 3.1 Data 3.2 BERTimbau 3.3 Self-Learning 4 Results and Discussion 5 Conclusion References Autonomous Learning Rate Optimization for Deep Learning 1 Introduction 2 Challenges and Constraints 3 Methods 3.1 Framing LR Control as a Learning Problem 3.2 Generating the Dataset 3.3 Correcting Ground Truth 3.4 Building the Model 4 Experiments 4.1 Image Classification on CIFAR10 4.2 Object Detection on MSCOCO 4.3 Language Modeling on PTB 5 Limitations and Unexpected Behaviors 6 Conclusion References Optimizing Data Augmentation Policy Through Random Unidimensional Search 1 Introduction 2 Methods 2.1 Dimensionality Reduction: 2D to 1D 2.2 More Search with Less Computation 2.3 RUA Augmentation Parameters 2.4 Selecting a Maximum N 3 Experiments 3.1 RUA Performance Assessment 3.2 Ablation Study 4 Conclusion References Evaluating Student Behaviour on the MathE Platform - Clustering Algorithms Approaches 1 Introduction 2 An Overview on Clustering Algorithms and Related Works 3 Clustering Approaches 3.1 Davies-Bouldin Index 3.2 Evolutionary Bio-inspired Clustering Algorithms 3.3 K-means Clustering Algorithm 4 Dataset 5 Results and Discussion 6 Conclusions References Unsupervised Training for Neural TSP Solver 1 Introduction 2 Related Work 3 Unsupervised TSP 3.1 Unsupervised Loss 3.2 Variable Discretization 3.3 Implementation 4 Neural Model 4.1 Graph Neural Network 5 Evaluation 6 Conclusions References Comparing Surrogate Models for Tuning Optimization Algorithms 1 Introduction 1.1 Related Work 2 Surrogate Models for Optimization Algorithms 3 Experimental Results 3.1 Methodology 3.2 Effects of Instance Representation and Pre-processing 3.3 Accuracy of the Surrogate Models 3.4 Agreement in Reproduction of Effects 4 Conclusions References Search and Score-Based Waterfall Auction Optimization 1 Introduction 2 Background and Related Works 3 Proposed Method 3.1 Estimate the Valuation Matrix 3.2 The Search Procedure 4 Empirical Evaluation 4.1 Synthetic Data 4.2 The Real-World Auction Data 5 Summary 6 Appendix References Survey on KNN Methods in Data Science 1 Introduction 2 Challenges 3 Choice of Distance Metric 4 Variations of KNN 5 Feature Selection and Data Reduction 6 Nearest Neighbor Matching Algorithms 7 Synopsis and Concluding Remarks References Constrained Shortest Path and Hierarchical Structures 1 Introduction 1.1 Our Contribution 2 Problem Formulation 3 Hierarchical Structures 3.1 Algorithm k-HSpmax 4 Algorithm A 5 Simulation 6 Conclusion References Investigation of Graph Neural Networks for Instance Segmentation of Industrial Point Cloud Data 1 Introduction 2 Background on Point Cloud Instance Segmentation Methods 2.1 Geometric Deep Learning Methods 2.2 Projection Based Methods 2.3 Graph Neural Networks 3 Experiments 3.1 Dataset 3.2 Implementation 4 Discussion 5 Conclusions References Fitness Landscape Ruggedness Impact on PSO in Dealing with Three Variants of the Travelling Salesman Problem 1 Introduction 2 Fitness Ruggedness 2.1 Fitness Landscape Features 2.2 Ruggedness Measures 3 Particle Swarm Optimization for TSP Landscapes 3.1 PSO for the TSP 3.2 Fitness Landscape Analysis for PSO 3.3 TSP Hardness and Fitness Landscape PSO Assessment 4 Experiments 4.1 Experimental Setup 4.2 Investigation of PSO Ruggedness on the Sets of Three Different TSP Variants 5 Conclusion and Discussion References A Multi-UAVs' Provider Model for the Provision of 5G Service Chains: A Game Theoretic Approach 1 Introduction 2 The Mathematical Formulation 3 Generalized Nash Equilibrium Problem Formulation 4 Illustrative Numerical Example 5 Conclusion and Future Works References Metabolic Syndrome Risk Forecasting on Elderly with ML Techniques 1 Introduction 2 Dataset Description 3 Data Preprocessing and Feature Importance 4 Performance Evaluation of ML Models 5 Conclusions References Airport Digital Twins for Resilient Disaster Management Response 1 Introduction 1.1 Background on Airport Resilience 1.2 Resilience Indexes 2 Airport Digital Twin Framework 2.1 Exploration of Threats and Hazards for Airport Digital Twins 2.2 Data Sources for the Generation and Maintenance of Airport DTs 3 Investigation of Environmental Digital Twin Metrics 4 Discussion 5 Conclusion References Strategies for Surviving Aggressive Multiparty Repeated Standoffs (Extended Abstract) 1 Introduction 1.1 Model, Notation, and Terminology 1.2 Related Work 1.3 Outline of the Paper 2 Observations on Standoffs 2.1 General Graphs 2.2 Dyadic (One-to-One) Duel 3 Standoffs on Rings 3.1 Standoffs on Unidirectional Rings 3.2 Triadic (Mexican) Standoff 4 Relationships 5 Standoffs on Complete Bipartite Graphs 5.1 One-to-Many Standoff 5.2 Many-to-Many Complete Bipartite Standoff 6 Conclusion References A Hybridization of GRASP and UTASTAR for Solving the Vehicle Routing Problem with Pickups and Deliveries and 3D Loading Constraints 1 Introduction 2 Vehicle Routing Problem with Pickups and Deliveries and 3D Loading Constraints 2.1 Problem Description 2.2 Solution Method 3 UTASTAR 4 Computational Experiments 4.1 Instances 4.2 Case Study 4.3 Computational Results 5 Conclusion References Packing Hypertrees and the k-cut Problem in Hypergraphs 1 Introduction 1.1 Previous Work 1.2 Our Contribution 1.3 Organization 2 Preliminaries 2.1 Separation of Partition Inequalities 2.2 Strength of a Network 2.3 Network Reinforcement 3 Packing Hypertrees 3.1 Integral Packing 3.2 Fractional Packing 4 A Relaxation of the k-cut Problem 4.1 Break-Points of l 4.2 The Maximum of l 4.3 An Upper Bound 5 A Linear Programming Relaxation for k-cut 6 A Polynomial Algorithm for Fixed and k 7 Concluding Remarks References Maximizing the Eigenvalue-Gap and Promoting Sparsity of Doubly Stochastic Matrices with PSO 1 Introduction 2 Preliminaries 2.1 Consensus and Eigenvalues 2.2 Relevant Approaches 2.3 Unified Particle Swarm Optimization 3 Problem Formulation 4 Experiments 5 Synopsis and Concluding Remarks References Value of Information in the Mean-Square Case and Its Application to the Analysis of Financial Time-Series Forecast 1 Introduction 2 Value of Information for Translation Invariant Objective Functions 3 Application: Analysis of Forecasts of Cryptocurrency Log-Returns 4 Discussion References Author Index