Evolutionary Multi-Criterion Optimization: 12th International Conference, EMO 2023, Leiden, The Netherlands, March 20–24, 2023, Proceedings

Description

This book constitutes the refereed proceedings of the 12th International Conference on Evolutionary Multi-Criterion Optimization, EMO 2022 held in Leiden, The Netherlands, during March 20-24, 2023. The 44 regular papers presented in this book were carefully reviewed and selected from 65 submissions. The papers are divided into the following topical sections: Algorithm Design and Engineering; Machine Learning and Multi-criterion Optimization; Benchmarking and Performance Assessment; Indicator Design and Complexity Analysis; Applications in Real World Domains; and Multi-Criteria Decision Making and Interactive Algorithms.. Preface Organization Contents Algorithm Design and Engineering Visual Exploration of the Effect of Constraint Handling in Multiobjective Optimization 1 Introduction 2 Background 2.1 Constrained Multiobjective Optimization Problems 2.2 Constraint Handling Techniques 3 Methodology 3.1 Test Problems 3.2 CMOP Landscape Visualization 3.3 Local Search 4 Experiments 4.1 Experimental Setup 4.2 Results and Discussion 5 Conclusions References A Two-Stage Algorithm for Integer Multiobjective Simulation Optimization 1 Introduction 2 Problem Definition 3 Algorithm 3.1 Pareto Retrospective Approximation Method for the First Stage 3.2 Local Stochastic Search for the Second Stage 4 Experimental Studies 4.1 Experiments on Test Instances 4.2 Biobjective Bus Scheduling 5 Conclusion References RegEMO: Sacrificing Pareto-Optimality for Regularity in Multi-objective Problem-Solving 1 Introduction 2 Motivation for Proposed Study 3 Past Studies 4 Regularity-Based Optimization (RegEMO) Procedure 4.1 Steps of Proposed RegEMO Procedure 5 Results and Discussion 5.1 Test Problems 5.2 Engineering Problems 6 Conclusions References Cooperative Coevolutionary NSGA-II with Linkage Measurement Minimization for Large-Scale Multi-objective Optimization 1 Introduction 2 Preliminaries and Related Works 2.1 Preliminaries 2.2 Decomposition Methods in LSMOPs 3 CC-NSGA-LMM 4 Numerical Experiments 4.1 Experiment Settings 4.2 Performance of CC-NSGA-LMM 4.3 Analysis 5 Discussion 5.1 Self-adaptation of Weight in LMF 5.2 More Powerful MOEAs 5.3 The Scalability of CC-NSGA-LMM 6 Conclusion References Data-Driven Evolutionary Multi-objective Optimization Based on Multiple-Gradient Descent for Disconnected Pareto Fronts 1 Introduction 2 Preliminaries 2.1 Basic Definitions in Multi-objective Optimization 2.2 Gaussian Process Regression Model 3 Proposed Method 3.1 MGD-Based Evolutionary Search 3.2 Infill Criterion 4 Experimental Setup 4.1 Benchmark Test Problems 4.2 Peer Algorithms and Parameter Settings 4.3 Performance Metric and Statistical Tests 5 Experimental Results 5.1 Performance Comparisons with the Peer Algorithms 5.2 Ablation Study 6 Conclusion References Eliminating Non-dominated Sorting from NSGA-III 1 Introduction 2 Proposed Algorithm: NGA-III(NSGA-III"026E30F NDS) 2.1 Classification of Pop. Members 2.2 Association of Population Members 2.3 Class-wise Mating Selection 2.4 Reference Vector Based Niching in Survival Selection 3 Experimental Results 3.1 Unconstrained Problems 3.2 Constrained Problems 4 Conclusions References Scalability of Multi-objective Evolutionary Algorithms for Solving Real-World Complex Optimization Problems 1 Introduction 2 Related Work with Objectives Reduction 3 Approach Based on Data Mining 3.1 Data Mining Methodology Adopted - FS-OPA 3.2 Comparison of FS-OPA with NL-MVU-PCA for MaOPs Data-Driven Structural Learning 4 Case Study 5 Results and Discussion 6 Conclusions References Machine Learning and Multi-criterion Optimization Multi-objective Learning Using HV Maximization 1 Introduction 2 Related Work 3 Approach 3.1 HV Maximization of Domination-Ranked Fronts 3.2 Implementation 3.3 A Toy Example 4 Experiments 4.1 MO Regression 4.2 Neural Style Transfer 5 Discussion References Sparse Adversarial Attack via Bi-objective Optimization 1 Introduction 2 Preliminaries 2.1 Adversarial Attack 2.2 Related Works 3 Proposed Method 3.1 Solution Evaluation 3.2 Offspring Population Generation 3.3 l2 Minimization Leads to l0 Minimization 4 Experimental Setup 4.1 Attack Setting 4.2 Performance Metric 4.3 Parameter Settings 5 Experimental Results 6 Conclusion and Future Directions References Investigating Innovized Progress Operators with Different Machine Learning Methods 1 Introduction 2 Background 3 Alternative ML Methods for IP2 and IP3 Operators 4 Experimental Settings 4.1 Test Suite and the Base RV-EMâOA 4.2 Performance Indicator 5 Experimental Results 6 Conclusion References End-to-End Pareto Set Prediction with Graph Neural Networks for Multi-objective Facility Location 1 Introduction 2 Related Work 2.1 Facility Location Problem 2.2 Graph Representation Learning 2.3 Machine Learning for Combinatorial Optimization on Graphs 3 Problem Formulation 4 Method 4.1 Bipartite Optimization on MO-FLP 4.2 The Dual GCN-Based Model 5 Experiments 5.1 Dataset Generation and Hyperparameter Configurations 5.2 Experimental Results 5.3 Hyperparameter Sensitivity Analysis 6 Conclusion and Future Work References Online Learning Hyper-Heuristics in Multi-Objective Evolutionary Algorithms 1 Introduction 2 Background 3 Selection Mechanisms: Single Selection and Distribution 3.1 Reward Function 3.2 Selection Pool 3.3 Selection Mechanism 4 Evaluation and Experiments 4.1 IGD Results of NSGA-II Using HHX-D, HHX-S, UX and SBX 4.2 Selection Behaviour of HHX-D and HHX-S 5 Advanced Selection Mechanisms: Evolving and Alternating 6 Comparison of All Presented Algorithms 7 Conclusion References Surrogate-assisted Multi-objective Optimization via Genetic Programming Based Symbolic Regression 1 Introduction 2 Surrogate-Assisted Multi-objective Optimization 2.1 Multi-objective Optimization Problem 2.2 Framework of Surrogate-Assisted Multi-objective Optimization 3 Surrogate Models 3.1 Kriging 3.2 Genetic Programming Based Symbolic Regression 3.3 GP-Based Symbolic Regression with Kriging 4 Acquisition Functions 5 Experiments 5.1 Parameter Settings 5.2 Empirical Results 6 Conclusion and Future Work References Learning to Predict Pareto-Optimal Solutions from Pseudo-weights 1 Introduction 2 Existing Studies 3 Proposed Machine Learning Based EMO Procedure 3.1 Training of Deep Neural Networks 3.2 Handling Variable Bounds and Constraints 4 Results 4.1 Two-Objective Problems 4.2 Three-Objective Problems 4.3 Many-Objective Optimization Problems 5 Conclusions References A Relation Surrogate Model for Expensive Multiobjective Continuous and Combinatorial Optimization 1 Introduction 2 Preliminaries 2.1 Problem Definition 2.2 Relation Learning 3 Proposed Method 3.1 Data Preparation 3.2 Model Training 3.3 Model Usage 4 Experimental Studies 4.1 Experimental Settings 4.2 Study on Continuous Problems 4.3 Study on Combinatorial Problems 5 Conclusion References Pareto Front Upconvert by Iterative Estimation Modeling and Solution Sampling 1 Introduction 2 Multi-objective Optimization 3 Pareto Front and Pareto Set Estimation 3.1 Overview 3.2 Pareto Front Estimation 3.3 Pareto Set Estimation 3.4 Estimated Set 4 Conventional SMOA 5 Proposal: Iterative SMOA (I-SMOA) 6 Experimental Settings 7 Experimental Results and Discussion 7.1 Obtained Upconverted Solution Set 7.2 Transition of Iterative Solution Insertion in I-SMOA 8 Conclusions References An Improved Fuzzy Classifier-Based Evolutionary Algorithm for Expensive Multiobjective Optimization Problems with Complicated Pareto Sets 1 Introduction 2 Related Work 2.1 Multiobjective Optimization Problems 2.2 Surrogate-Based MOEAs 3 Our Proposed Algorithm 3.1 Algorithm Framework 3.2 IFCS-Based Sorting 4 Experiments 4.1 Experimental Settings 4.2 Effect of IFCS on MOEA/D-DE 4.3 Performance Comparison with the State-of-the-art MOEAs 4.4 Performance Comparison on Real-World Problems 5 Conclusion References Approximation of a Pareto Set Segment Using a Linear Model with Sharing Variables 1 Introduction 2 Performance Metric for Local Models 2.1 Local Approximation Metric 2.2 Shared Variable Metric 3 Linear Sparse Representation of the Local PS 3.1 Linear Model 3.2 Variable Sharing and Sparsity of the Model 4 Method and Algorithm 4.1 An Alternative Problem 4.2 Algorithm Framework 5 Experimental Results 5.1 Parameter Setting 5.2 Performance on None-shared Problem 5.3 Performance on Standard Test Instances 6 Conclusion References Feature-Based Benchmarking of Distance-Based Multi/Many-objective Optimisation Problems: A Machine Learning Perspective 1 Introduction 2 Distance-based Multi/Many-objective Problems 3 Experimental Setup 3.1 Dataset 3.2 Algorithm Performance 4 Experimental Study 4.1 Problem Features vs Algorithm Performance 4.2 Performance Prediction by Regression 4.3 Algorithm Selection by Classification 5 Conclusions References Benchmarking and Performance Assessment Partially Degenerate Multi-objective Test Problems 1 Introduction 2 DTLZ5, DTLZ6 and WFG3 Test Problems 2.1 Pareto Fronts of DTLZ5, DTLZ6 and WFG3 2.2 Availability of the Test Problems 3 Performance Evaluation Results 4 Conclusions References Peak-A-Boo! Generating Multi-objective Multiple Peaks Benchmark Problems with Precise Pareto Sets 1 Introduction 2 Background 3 On the Pareto Set of Multiple Peaks Functions 4 Experimental Study 5 Conclusions References MACO: A Real-World Inspired Benchmark for Multi-objective Evolutionary Algorithms 1 Introduction 2 Related Works 3 Multi-agent Coordination Problem 3.1 Variation: P-Norm 3.2 Variation: Weights 3.3 Variation: Interaction Classes 3.4 Optimal Solution 4 Experiments 5 Conclusion References A Scalable Test Suite for Bi-objective Multidisciplinary Optimization 1 Introduction 2 Related Literature 2.1 Multi-disciplinary Benchmarks 2.2 Multi-objective Benchmarks 2.3 Multi-objective Multidisciplinary Benchmarks 3 Proposed MO-MDO Test Suite: ZDT-MDO 4 Defining Dependencies Between Disciplines 5 Experimental Setup 6 Experimental Results 7 Summary and Future Work References Performance Evaluation of Multi-objective Evolutionary Algorithms Using Artificial and Real-world Problems 1 Introduction 2 Artificial Test Problems and Real-World Problems 3 Results of Computational Experiments 4 Concluding Remarks References A Novel Performance Indicator Based on the Linear Assignment Problem 1 Introduction 2 Background 2.1 Multi-objective Optimization 2.2 Linear Assignment Problem 3 Our Proposed Indicator 4 Comparison Between Our Approach and the R2-indicator 5 Experimental Analysis 5.1 Evaluation in Artificial Many-Objective Pareto Fronts 5.2 Evaluation in Pareto Front Approximations 6 Conclusions and Future Work References A Test Suite for Multi-objective Multi-fidelity Optimization 1 Background 2 The Proposed MOMF Problem Suite 3 Numerical Experiments and Discussion 4 Conclusion and Future Work References Indicator Design and Complexity Analysis Diversity Enhancement via Magnitude 1 Introduction 2 Weightings, Magnitude, and Diversity 3 The Weighting Gradient Flow 4 Enhancing Diversity 5 Performance on Benchmarks 6 Algorithmic Extensions 6.1 Multi-objective Weighting Gradient Flow 6.2 Recycling Function Evaluations 7 Remarks References Two-Stage Greedy Approximated Hypervolume Subset Selection for Large-Scale Problems 1 Introduction 2 Greedy Hypervolume Subset Selection 2.1 Greedy Exact HSS Methods 2.2 Greedy Approximated HSS Method 3 Proposed Two-Stage Greedy Approximated HSS 4 Experimental Results 4.1 Experimental Settings 4.2 Performance of TGAHSS Under Different Parameter Settings 4.3 Comparison with State-of-the-Art Methods 5 Conclusion References The Hypervolume Indicator Hessian Matrix: Analytical Expression, Computational Time Complexity, and Sparsity 1 Introduction 2 General Construction of Hypervolume Hessian and Gradient via the Chain Rule 3 Hypervolume Indicator Hessian Matrix in 3-D 4 General N-Dimensional Expression of the Hypervolume Hessian Matrix 4.1 Partial Derivative (H/yk(i))/ y (i) 4.2 Partial Derivative (H/yk(i))/ y (j), i=j 5 Numerical Examples 6 Discussion and Outlook References On the Computational Complexity of Efficient Non-dominated Sort Using Binary Search 1 Introduction 2 Approach 3 Best Case Analysis of ENS-BS 4 Identified Scenario 4.1 Dominance Comparisons Between Points of the Same Front 4.2 Dominance Comparisons Between Points of Different Fronts 5 Conclusion and Future Work References Applications in Real World Domains Evolutionary Algorithms with Machine Learning Models for Multiobjective Optimization in Epidemics Control 1 Introduction 2 Optimization Problem 2.1 Simulation-Based Calculation of the Objective f2 2.2 Real-Life Attributes Affecting the Susceptibility to the Disease 2.3 Modelling Disease Transmission Probability Using Graph Node Attributes 3 Optimization Algorithms 3.1 EA-C 3.2 EA-R 3.3 MOEA/D 4 Experiments and Results 5 Conclusion References Joint Price Optimization Across a Portfolio of Fashion E-Commerce Products 1 Introduction 2 Related Work 3 Methodology 3.1 Optimization 3.2 Product Clusters 3.3 Demand Model 3.4 Overall Price Recommendation Pipeline 4 Offline Demand Model Evaluation 4.1 Data Preparation 5 Comparing GA Solutions to MILP Solutions 6 Online Evaluation 6.1 Test Framework 6.2 Results 7 Conclusion and Future Work References Improving MOEA/D with Knowledge Discovery. Application to a Bi-objective Routing Problem 1 Introduction 2 Scientific Context 2.1 Knowledge Discovery in Metaheuristics 2.2 Knowledge Integration in Multi-objective Optimization 3 Knowledge Discovery for Multi-objective Optimization 3.1 Definition of Knowledge Groups 3.2 Intensification and Diversification Strategies 4 MOEA/D Enhanced with Knowledge Discovery 4.1 MOEA/D 4.2 Construction of the Knowledge Groups and Strategies 4.3 MOEA/D with Knowledge Discovery 4.4 Experimental Variants 5 Bi-Objective Vehicle Routing Problem with Time Windows (bVRPTW) 5.1 Problem Description 5.2 Related Works 5.3 Local Search and Knowledge Operators 6 Experimental Setup 6.1 Solomon's Benchmark 6.2 Termination Criterion and Performance Assessment 6.3 Tuning 7 Experimental Protocol 8 Experimental Results and Discussion 9 Conclusion References The Prism-Net Search Space Representation for Multi-objective Building Spatial Design 1 Introduction 2 Search Space Representation 3 Mutation Operators 4 Integration to NSGA-II and SMS-EMOA 5 Summary and Outlook References Selection Strategies for a Balanced Multi- or Many-Objective Molecular Optimization and Genetic Diversity: A Comparative Study 1 Introduction 2 Preliminary Work 3 Proposed Approach and ad-MOEA 4 Experimental Setup 4.1 Molecular Optimization Problems 4.2 Performance Metrics 4.3 Experimental Results 5 Conclusion References A Multi-objective Evolutionary Framework for Identifying Dengue Stage-Specific Differentially Co-expressed and Functionally Enriched Gene Modules 1 Introduction 2 Methods 2.1 Computation of Co-expression Similarity of Genes 2.2 Computation of Differential Similarity of Genes for Different Disease States 2.3 Computation of Semantic Similarity of Genes 2.4 Differentially Co-expressed Module Identification 2.5 Most Promising Gene Identification 2.6 Statistical Significance of the Identified Modules 2.7 Comparison with Existing Methods 3 Biological Significance Study of the Identified Modules 3.1 Significance of the Identified Markers 4 Conclusion References Real-World Airline Crew Pairing Optimization: Customized Genetic Algorithm Versus Column Generation Method 1 Introduction 2 Airline Crew Pairing Optimization Problem 3 Genetic Algorithm 3.1 Novel Chromosome Representation 3.2 Fitness Evaluation 3.3 Minimal-Deadhead Initialization Heuristic 3.4 Selection 3.5 Crossover 3.6 Mutation 3.7 Feasibility-Repair Heuristic 3.8 Population Replacement 4 Computational Experiments 5 Conclusion References Multiobjective Optimization of Evolutionary Neural Networks for Animal Trade Movements Prediction 1 Introduction 2 Prediction of Animal Transports 3 Neural Networks 4 Optimization Problem 5 Experiments and Results 5.1 Parameter Tuning 5.2 Testing the Generalization Capability of the Neural Models 6 Conclusion References Transfer of Multi-objectively Tuned CMA-ES Parameters to a Vehicle Dynamics Problem 1 Introduction 2 Covariance Matrix Adaptation Evolution Strategy 3 Real-World Problem Description 4 Exploratory Landscape Analysis for Transfer Learning 5 Multi-objective Tuning of Algorithm's Parameters on Reference Functions 5.1 Results 5.2 Transfer of Tuned Parameters to the Real-World Problem 6 Conclusion References Multi-criteria Decision Making and Interactive Algorithms Preference-Based Nonlinear Normalization for Multiobjective Optimization 1 Introduction 2 Preliminaries 2.1 Multiobjective Optimization Problem 2.2 Linear Normalization 2.3 Preference Incorporation 3 Proposed Preference-Based Nonlinear Normalization 4 Experimental Studies 4.1 Relation to Linear Normalization 4.2 Incorporation into Different MOEAs 4.3 Analysis in the Transformed Objective Space 5 Conclusion References Incorporating Preference Information Interactively in NSGA-III by the Adaptation of Reference Vectors 1 Introduction 2 Background 2.1 Multiobjective Optimization 2.2 Evolutionary Algorithms 3 Related Work 4 Proposal 5 Algorithmic Comparison 5.1 Discussion 6 Conclusions References A Systematic Way of Structuring Real-World Multiobjective Optimization Problems 1 Introduction 2 Background 2.1 Basic Concepts 2.2 Brief Literature Review 3 Systematic Way of Structuring MOO Problems 4 Discussion 5 Conclusions References IK-EMOViz: An Interactive Knowledge-Based Evolutionary Multi-objective Optimization Framework 1 Introduction 2 Proposed IK-EMO Visualizer (IK-EMOViz) 2.1 User-Provided Knowledge Before the Optimization 2.2 Automatic Knowledge-Extraction During Optimization 2.3 Knowledge Application Through Repair Operators 2.4 User Feedback Through IK-EMOViz's Graphical User Interface 2.5 Synchronous vs Asynchronous User Interaction 3 Truss Design Problem 3.1 Experimental Settings 3.2 Experimental Results and Discussion on the 120-Member Truss 3.3 Experimental Results and Discussion on the 820-Member Truss 4 Conclusions and Future Work References An Interactive Decision Tree-Based Evolutionary Multi-objective Algorithm 1 Introduction 2 Methods 2.1 Decision Tree-Based EMOA (DTEMOA) 3 Experimental Design 3.1 Machine Decision Maker (MDM) 3.2 Benchmark Problems 3.3 Evaluating Performance and Competing Algorithms 3.4 Algorithm Parameter Settings 4 Results and Discussion 4.1 Assessing Ranking Performance 4.2 Comparison of the Performance with Other iEMOAs 5 Conclusions References Author Index