Optimization and Learning: 4th International Conference, OLA 2021, Catania, Italy, June 21-23, 2021, Proceedings (Communications in Computer and Information Science)

This volume constitutes the refereed proceedings of the 4th International Conference on Optimization and Learning, OLA 2021, held in Catania, Italy, in June 2021. Due to the COVID-19 pandemic the conference was held online. The 27 full papers were carefully reviewed and selected from 62 submissions. The papers presented in the volume are organized in topical sections on ​synergies between optimization and learning; learning for optimization; machine learning and deep learning; transportation and logistics; optimization; applications of learning and optimization methods. Preface Organization Contents Synergies Between Optimization and Learning Embedding Simulated Annealing within Stochastic Gradient Descent 1 Introduction 2 Simulated Annealing 2.1 A Naive Implementation for Training Without Gradients 3 Improved SGD Training by SA 4 Computational Analysis of SGD-SA 5 Conclusions and Future Work References Comparing Local Search Initialization for K-Means and K-Medoids Clustering in a Planar Pareto Front, a Computational Study 1 Introduction 2 Problem Statement and Notations 3 Initialization Heuristics 3.1 Generic Initialization Strategies 3.2 Initialization Using 2D PF Indexation 3.3 Initialization Using p-Dispersion for 2D PF 3.4 Initialization Using 1D Dynamic Programming 4 Computational Experiments and Results 4.1 Data Generation 4.2 Computational Experiments and Conditions 4.3 Analyses of Computational Results 5 Conclusions and Perspectives References Reinforcement Learning-Based Adaptive Operator Selection 1 Introduction 2 Adaptive Operator Selection 2.1 Operator Selection 2.2 Credit Assignment 3 Proposed Approach: Adaptive Selection with Reinforced-Clusters 4 Experimental Results 5 Conclusion References Learning for Optimization A Learning-Based Iterated Local Search Algorithm for Solving the Traveling Salesman Problem 1 Introduction 2 Literature Review 3 Preliminaries 3.1 Traveling Salesman Problem 3.2 Iterated Local Search 3.3 The Q-Learning Algorithm 4 Proposed Q-ILS Algorithms 4.1 Q-ILS-1 Algorithm 4.2 Q-ILS-2 Algorithm 5 Results and Discussion 5.1 Experimental Design 5.2 Numerical Results 6 Conclusion References A Hybrid Approach for Data-Based Models Using a Least-Squares Regression 1 Introduction 2 Data-Based Modeling Approach 2.1 Data-Based Modeling with Least-Squares Regression 2.2 Introducing Hybrid Models 3 Experimental Results Comparing Non-hybrid and Hybrid Models on Real-World Data 3.1 Setup and Data 3.2 Comparing Non-hybrid and Hybrid Models on Real-World Data 3.3 Checking the Models' Plausibility 4 Conclusion References A Comparison of Learnheuristics Using Different Reward Functions to Solve the Set Covering Problem 1 Introduction 2 Set Covering Problem 3 Reinforcement Learning 3.1 Q-Learning 3.2 Reward Function 4 Sine Cosine Algorithm 4.1 Learnheuristic Framework 4.2 Balancing Exploration and Exploitation 5 Experimental Results 6 Conclusion References A Bayesian Optimisation Approach for Multidimensional Knapsack Problem 1 Introduction 2 Two-Level Model for MKP 3 Bayesian Optimisation and Acceleration 3.1 Variable Domain Tightening 3.2 Initialisation with Genetic Algorithm 3.3 Optimisation Landscape Smoothing 4 Implementation 5 Computational Experiments 6 Conclusion and Future Work References Machine Learning and Deep Learning Robustness of Adversarial Images Against Filters 1 Introduction 2 VGG-16 Trained on CIFAR-10 3 Target and Untargeted Scenarios, and Design of EAL2target 4 Obtention of the Adversarial Images: Running EAL2target 5 Selection of Filter 6 Filtering the Ancestor and the Adversarial Images 7 The Variant EAL2target, F 8 Conclusion References Guiding Representation Learning in Deep Generative Models with Policy Gradients 1 Introduction 2 Related Work 3 Combination of Reinforcement and Representation Learning Objectives 3.1 Reinforcement Learning with Policy Optimization 3.2 Learning Representations Using Variational Auto-Encoders 3.3 Joint Objective Function 4 Experiments 4.1 Data Collection and Pre-Processing 4.2 Pre-training the Variational Auto-Encoder 4.3 Jointly Learning Representation and Policy 4.4 Analyzing the Value Function Gradients 5 Conclusion A Appendix A.1 Stable Policy Learning with Proximal Policy Optimization A.2 Hyperparameter Tables A.3 Choosing Appropriate Values for References Deep Reinforcement Learning for Dynamic Pricing of Perishable Products 1 Introduction 2 Related Work 3 MDP Formulation for Dynamic Pricing of Perishables 4 Methodology 5 Experimental Results 6 Conclusion References An Exploratory Analysis on a Disinformation Dataset 1 Introduction 2 Related Work 3 Background Theory 3.1 Hierarchical Clustering 3.2 t-SNE 4 Methodology 4.1 Fake.br Corpus 4.2 Pre-processing 4.3 Agglomerative Clustering 4.4 Brazilian Disinformation Corpus 4.5 Outsiders Analysis 5 Results 6 Conclusion References Automatic Synthesis of Boolean Networks from Biological Knowledge and Data 1 Introduction 2 Boolean Networks and Their Synthesis 2.1 Prior Knowledge Network (PKN) 2.2 Boolean Networks (BNs) 2.3 Synthesis of BNs from PKN and Multivariate TS 3 State-of-the-Art Methods of BN Synthesis from PKN and TS 4 Our Approach: ASKeD-BN 4.1 Details of the Approach 4.2 Illustration on the Toy Example 5 Datasets and Procedure for the Comparative Evaluation 5.1 Datasets 5.2 Details on the Evaluation Procedure 6 Results 6.1 Results on Systems with Real PKN and Experimental Multivariate TS 6.2 Results on Systems with Generated Multivariate TS 7 Conclusion and Perspectives References Transportation and Logistics Solving Inventory Routing Problems with the Gurobi Branch-and-Cut Algorithm 1 Introduction 2 Literature Survey 3 Model 3.1 Problem Description and Formulation 3.2 Mathematical Formulation 3.3 Test Instances 4 Algorithm 4.1 Implementation of Subtour Elimination Constraints 5 Results 5.1 Algorithm Optimization Results 5.2 Computation Results 6 Conclusions References Iterated Local Search with Neighbourhood Reduction for the Pickups and Deliveries Problem Arising in Retail Industry 1 Introduction 2 Problem Statement 3 ILS with Neighbourhood Reduction 3.1 INITIAL Procedure 3.2 PERTURB Procedure 4 Computational Study 5 Conclusion References A Genetic Algorithm for the Three-Dimensional Open Dimension Packing Problem 1 Introduction 2 Problem Formulation 3 Genetic Algorithm 3.1 Solution Encoding and Decoding 3.2 Solution Construction 3.3 Evolutionary Process 4 Computational Experiments 5 Conclusion References Formulation of a Layout-Agnostic Order Batching Problem 1 Introduction 2 Literature Review 3 Problem Formulation 3.1 Preliminaries 3.2 General OBP Formulation 3.3 Single Batch OBP Formulation 4 Experimental Results 5 Conclusion References Optimization Neighborhood Enumeration in Local Search Metaheuristics 1 Introduction 2 Iterative Improvement, Neighborhoods and Selection 3 Neighborhood Enumeration 3.1 Parameter Spaces 3.2 Enumeration Order 4 Experimental Evaluation 5 Discussion 6 Conclusion References Cryptographic Primitives Optimization Based on the Concepts of the Residue Number System and Finite Ring Neural Network 1 Introduction 2 Modular Arithmetical Operations 2.1 Addition, Subtraction, Multiplication, and Division 2.2 Euclidean Division 3 Modular Logical Operations 4 Scaling RNS Numbers by Base Extension 5 Experimental Results 6 Conclusion References Investigating Overlapped Strategies to Solve Overlapping Problems in a Cooperative Co-evolutionary Framework 1 Introduction 2 Related Work 2.1 Cooperative Co-evolutionary Algorithms 2.2 Recursive Differential Grouping 3 Proposed Algorithm 3.1 Overlapped Recursive Differential Grouping 3.2 Overlapped CC Framework 4 Experimental Settings and Results 5 Discussion References Improved SAT Models for NFA Learning 1 Introduction 2 Modeling the Problem in SAT 3 Improving the SAT Model 4 Experimental Results 5 Conclusion References Applications of Learning and Optimization Methods Synthesis of Scheduling Heuristics by Composition and Recombination 1 Introduction 2 Classification of Machine Scheduling Problems 3 Related Work 3.1 Machine Scheduling Algorithms for Flow Shops and Job Shops 3.2 Giffler & Thompson Algorithm 3.3 Combinatory Logic Synthesizer 4 Implementation 5 Results 6 Conclusion References Solving QAP with Auto-parameterization in Parallel Hybrid Metaheuristics 1 Introduction 2 Related Work 3 DPA-QAP Method 3.1 Metaheuristics Used in the DPA-QAP Method 4 Automatic Parameter Adaption in DPA-QAP 4.1 Metaheuristics Performance Metrics 4.2 Performance Evaluation 5 Experimental Evaluation 5.1 Evaluation on QAPLIB 5.2 Evaluation on Harder Instances 6 Conclusions and Future Work References Theoretical Analysis of a Dynamic Pricing Problem with Linear and Isoelastic Demand Functions 1 Introduction 2 Problem Description 3 Resolution Approach 3.1 Case with Linear Demand Function 3.2 Case with Isoelastic Demand Function 4 Numerical Experiments 5 Conclusion References A Hybrid FLP-AHP Approach for Optimal Product Mix in Pulp and Paper Industry 1 Introduction 2 Literature Review 3 Methodology 4 Case Study 4.1 Background 5 Mathematical Model 5.1 Formulation of FLP 5.2 AHP 6 Results and Discussions 6.1 Sensitivity Analysis 6.2 Decision Making Through AHP 7 Conclusion References An Application of BnB-NSGAII: Initializing NSGAII to Solve 3 Stage Reducer Problem 1 Introduction 2 3 Stage Reducer Problem 3 BnB-NSGAII 3.1 General Concept of BnB-NSGAII 3.2 BnB Legacy Feature 3.3 An Application of BnB-NSGAII 4 Numerical Experiment 4.1 Results and Discussion 5 Conclusion A 3SR Problem Constraints A.1 Closure Condition A.2 Mechanical Constraint for One Stage of the Mechanism References The Horizontal Linear Complementarity Problem and Robustness of the Related Matrix Classes 1 Introduction 2 Particular Matrix Classes 2.1 S-matrices 2.2 Nonnegative Matrices 2.3 Z-matrices 2.4 Semimonotone Matrices 2.5 Principally Nondegenerate Matrices 2.6 Column Sufficient Matrices 2.7 R0-matrices 2.8 R-matrices 3 Conclusion References Incorporating User Preferences in Multi-objective Feature Selection in Software Product Lines Using Multi-Criteria Decision Analysis 1 Introduction 2 Background 2.1 Software Product Line Engineering 2.2 Multi-Objective Optimisation 3 State-of-the-Art and Proposed Approach 3.1 SATIBEA 3.2 Multi-Criteria Decision Analysis 4 System Set-Up 4.1 System and Algorithms Set-Up 4.2 Dataset 4.3 Multi-objective Performance Metrics 5 Evaluation 5.1 Execution Time Overhead 5.2 Multi-objective Performance Metrics 5.3 SAT_MCDA_EA's Strictly Non-Dominated Solutions 6 Conclusion and Future Work References Author Index