Computational Science – ICCS 2022: 22nd International Conference, London, UK, June 21–23, 2022, Proceedings, Part I (Lecture Notes in Computer Science, 13350)

The four-volume set LNCS 13350, 13351, 13352, and 13353 constitutes the proceedings of the 22ndt International Conference on Computational Science, ICCS 2022, held in London, UK, in June 2022.\*The total of 175 full papers and 78 short papers presented in this book set were carefully reviewed and selected from 474 submissions. 169 full and 36 short papers were accepted to the main track; 120 full and 42 short papers were accepted to the workshops/ thematic tracks. \*The conference was held in a hybrid format ICCS 2022 Main Track Full Papers Developing a Scalable Cellular Automaton Model of 3D Tumor Growth 1 Introduction 2 Related Work 3 Methods 3.1 Extension of 2D Cellular Automaton Model to 3D 3.2 N-body and Lattice Parallelization Schemes 3.3 Adaptive Communication Scheme to Reduce Overhead 3.4 3D Serial and Parallel Verification Protocols 3.5 Performance Evaluation of the Parallelized 3D Model 4 Results 4.1 3D Model Preserves Per-Slice Tumor Morphology 4.2 3D Parallel Model Matches 3D Serial Model 4.3 Strong Scaling Indicates When to Launch the Parallel Model 4.4 Throughput and Efficiency Increases over Time 5 Discussion and Conclusion References Human-Level Melodic Line Harmonization 1 Introduction 1.1 Contribution 2 Problem Definition 3 Related Literature 4 Proposed Method 4.1 Problem Search Space - Admissible Chords 4.2 Population Generation Process 4.3 Selection Method 4.4 Mutation and Crossover 4.5 Fitness Function 5 Experimental Setup and Results 5.1 Parameterization 5.2 Efficacy of the Algorithm - Formal Aspects 5.3 Human Expert Evaluation Including Aesthetic Aspects 5.4 Running Time 5.5 Modeling the Solution 6 Conclusions and Future Work References Classification Methods Based on Fitting Logistic Regression to Positive and Unlabeled Data 1 Introduction 2 Objectives and Methods 3 Numerical Experiments 3.1 Datasets 3.2 Results 4 Conclusions References Robust Control of Perishable Inventory with Uncertain Lead Time Using Neural Networks and Genetic Algorithm 1 Introduction 2 Problem Definition and Assumptions 3 Proposed Approach 4 Simulation Results 4.1 Lead-Time Influence on Testing Error 4.2 Lead-Time Perturbation Influence on the Learning Process 4.3 Robustness of Proposed Approach 4.4 Time Responses of the Obtained NNC Controllers for Long-Lead-Time Scenario 4.5 Test Size Influence on Testing Error 5 Conclusions References Batch QR Factorization on GPUs: Design, Optimization, and Tuning 1 Introduction and Related Work 2 Algorithmic Background 2.1 Nested Blocking 2.2 Computing the Triangular Factor T 3 Experimental Setup 4 LAPACK-Style Design 5 Panel Optimization 6 Optimizing the Trailing Matrix Update 7 Optimizations for Sub-warp Dimensions 8 The Big Picture 9 Final Performance Results 10 Conclusion and Future Work References ChemTab: A Physics Guided Chemistry Modeling Framework 1 Introduction 2 Related Work 3 ChemTab: Joint Learning Progress Variables and Manifold Approximation 3.1 Background: Unsteady FGM 3.2 ChemTab 4 Experimentation and Results 4.1 Dataset 4.2 Evaluation 4.3 Implementation and Settings 4.4 Compared Methods 4.5 Results 5 Conclusion References Establishing Metrics to Quantify Underlying Structure in Vascular Red Blood Cell Distributions 1 Introduction 2 Methods and Metrics 2.1 Generating Initial Configurations of Red Blood Cells in a Microvessel 2.2 Quantifying Spatial Similarity Between Cell Configurations 3 Results and Discussion 3.1 Applying the Radial Distribution Function to Quantify Structure in a Single Packing 3.2 Utilizing Jaccard Index to Quantify Distributions of Cell Configurations 4 Conclusion References Coevolutionary Approach to Sequential Stackelberg Security Games 1 Introduction 2 Problem Definition 3 Related Work 4 Coevolutionary Approach 5 Experimental Setup 5.1 Benchmark Games 5.2 Parameterization 6 Results and Discussion 7 Conclusions References Finch: Domain Specific Language and Code Generation for Finite Element and Finite Volume in Julia 1 Introduction 2 Related Work 3 Domain Specific Language 3.1 User Input 3.2 Symbolic Representation 4 Code Generation 4.1 Elemental Computation 4.2 Global Computation 4.3 Modifying Generated Code 5 Performance Opportunities 5.1 Performance Within Finch 5.2 Cache Optimization 6 Demonstration 6.1 Steady-State Advection-Diffusion-Reaction Equation 7 Conclusion References Classifying Anomalous Members in a Collection of Multivariate Time Series Data Using Large Deviations Principle: An Application to COVID-19 Data 1 Introduction 2 Related Work 3 Large Deviation Principle 4 Methodology 4.1 Large Deviations for Anomaly Detection 4.2 LDP for High Dimensional Data 4.3 LAD for Time Series Data 5 Experiments 5.1 Datasets 5.2 Baseline Methods and Parameter Initialization 5.3 Evaluation Metrics 5.4 Anomaly Detection Performance 5.5 Anomaly Detection in Individual Time Series 5.6 Anomaly Detection in Time Series Data 5.7 Discoveries: US COVID-19 Trends 5.8 Global Trends and Emergence of Other COVID-19 Variants 6 Conclusion References Adaptive Regularization of B-Spline Models for Scientific Data 1 Introduction 2 Related Work 3 Background on B-Splines 3.1 B-Spline Curves 3.2 Tensor Product B-Splines 3.3 Optimal Control Points 4 Adaptive Regularization 5 Results 5.1 Data Sets 5.2 Comparison of Adaptive Versus Uniform Regularization 5.3 Accuracy on Analytical Signals 5.4 Extrapolation into Unconstrained Regions 6 Future Work 7 Conclusions References Content-Aware Generative Model for Multi-item Outfit Recommendation 1 Introduction 2 Related Works 2.1 Recommender-Based Solutions 2.2 GAN-Based Solutions 3 Transformer-Based GAN for Outfit Recommendations 3.1 Problem Formulation 3.2 Multimodal Transformer-Based GAN with Cross-Modal Attention 4 Experiments 4.1 Dataset 4.2 Baselines 4.3 Evaluation and Results 5 Conclusions References Retrofitting Structural Graph Embeddings with Node Attribute Information 1 Introduction 2 Related Work 2.1 Structural Representation Learning Methods 2.2 Attributed Graph Embedding Methods 3 Notations and Problem Definition 4 Graph Embedding RetroFitting (GERF) 4.1 Objective Function 4.2 Optimization 4.3 Summary 5 Experimental Setup 5.1 Datasets 5.2 Embedding Methods 5.3 Baselines 5.4 GERF Hyperparameters 6 Node Classification 7 Conclusions References Hierarchical Ensemble Based Imbalance Classification 1 Introduction 2 Related Work 3 The Proposed Method 3.1 Bootstrap Sampling 3.2 Instance Selection and Weighting Strategy 3.3 Multivariate Gaussian Distribution Based Oversampling 3.4 Hierarchical Ensemble 4 Datasets and Experiment Setting 4.1 Evaluation Metrics 5 Experiments 5.1 The Effect of Hyperparameters 5.2 Our Method vs. Undersampling-Based Ensembles 5.3 Our Method vs. Oversampling-based Ensembles 6 Conclusion and Future Work References Numerical Approximation of the One-Way Helmholtz Equation Using the Differential Evolution Method 1 Introduction 2 Rational Approximation of the One-Way Helmholtz Equation 3 Dispersion Relation 4 Optimization of the Numerical Scheme 4.1 Differential Evolution Method 4.2 Unconditional Optimization 4.3 Optimization for a Given Accuracy 4.4 Stability Condition 5 Numerical Results 6 Conclusion References Iterative Solution for the Narrow Passage Problem in Motion Planning 1 Introduction 2 Related Work 3 Proposed Solution 4 Experiments and Results 5 Conclusion References A Productive and Scalable Actor-Based Programming System for PGAS Applications 1 Introduction 2 Background: Communication in PGAS Applications 2.1 Read Pattern 2.2 Update Pattern 2.3 Conveyors 3 Our Approach 3.1 Fine-grained-Asynchronous Bulk-Synchronous Parallelism (FA-BSP) Model 3.2 High-Level Design of Programming System 3.3 User-facing API 3.4 Class-based API 4 Implementation 4.1 HClib Asynchronous Many-Task Runtime 4.2 Execution Model 4.3 Selector Runtime 4.4 Source-to-Source Translation from Lambda-Based to Class-Based Messaging 5 Evaluation 6 Related Work 7 Conclusions and Future Work References Is Context All You Need? Non-contextual vs Contextual Multiword Expressions Detection 1 Introduction 2 Related Work 3 Dataset 3.1 plWordNet-based Non-Contextual Dataset 3.2 KGR10-based Contextual Dataset 4 Methods for Multiword Expression Detection 5 Experiments 6 Results 7 Discussion 8 Conclusions and Future Work References Out-of-Distribution Detection in High-Dimensional Data Using Mahalanobis Distance - Critical Analysis 1 Introduction 2 OoD Detection with Mahalanobis Distance - Simulation Study 2.1 Method - Using Mahalanobis Distance for OoD Detection 2.2 Non-robust Estimation of MVN Model in High Dimensional Data 2.3 Non-robust MVN Model Used for OoD Detection 2.4 Mahalanobis Distance-based vs. Nonparametric Outlierness Factor-based OoD in High Dimensional Data 2.5 Evaluation Metrics 3 Using Mahalanobis Distance for OoD Detection in CNNs 3.1 Datasets and CNN Models 3.2 Analysis of Mahalonobis Distances for the CIFAR-10 3.3 LOF-based OoD 3.4 OoD Experiments 4 Conclusion References Multi-contextual Recommender Using 3D Latent Factor Models and Online Tensor Decomposition 1 Introduction 2 Related Work 3 Methodology 4 Results 5 Discussion References Efficient Computational Algorithm for Stress Analysis in Hydro-Sediment-Morphodynamic Models 1 Introduction 2 Equations for Hydro-Sediment-Morphodynamic Models 3 A Coupled Finite Element/Finite Volume Method 3.1 Well-balanced Finite Volume Solution of Sediment Transport 3.2 Unstructured Finite Element Solution of Elasticity 3.3 Implementation of Coupling Conditions at the Interface 4 Numerical Results 5 Concluding Remarks References Enhancing Computational Steel Solidification by a Nonlinear Transient Thermal Model 1 Introduction 2 Mathematical Models for Steel Solidification 3 Numerical Methods for the Nonlinear System 4 Results and Discussions 4.1 Accuracy Example 4.2 Steel Solidification: Case 1 4.3 Steel Solidification: Case 2 5 Conclusions References Weakly-Supervised Cell Classification for Effective High Content Screening 1 Introduction 2 Related Works 2.1 High Content Screening 2.2 Weakly-Supervised Learning 2.3 Self-supervised Learning 3 Data 3.1 Data Acquisition 3.2 Data Labelling 4 Methods 4.1 Position-Aware Self-Supervised Multiple Instance Learning Method (SSMIL) 5 Experimental Setup 6 Results 7 Conclusions References Which Visual Features Impact the Performance of Target Task in Self-supervised Learning? 1 Introduction 2 Related Works 3 Methods 4 User Studies 5 Experimental Setup 6 Results 7 Conclusions References DITA-NCG: Detecting Information Theft Attack Based on Node Communication Graph 1 Introduction 2 Related Work 3 Modeling and Methodology 3.1 Data Pre-process 3.2 Attack Detection 4 Experiment and Evaluation 4.1 Experimental Data 4.2 Experimental Result and Analysis 4.3 Comparison with Other Methods 5 Conclusion and Discussion References Boosted Ensemble Learning Based on Randomized NNs for Time Series Forecasting 1 Introduction 2 Base Forecasting Model - RandNN 3 Boosting of Ensemble Learning 3.1 Ensemble Learning Based on Residuals 3.2 Ensemble Learning Based on Corrected Targets 3.3 Ensemble Learning Based on Opposed Response 4 Experimental Study 5 Conclusion References Exploring Ductal Carcinoma In-Situ to Invasive Ductal Carcinoma Transitions Using Energy Minimization Principles 1 Introduction 2 Methods and Materials 2.1 Model Description 3 Results and Discussions 4 Conclusions References Scaling the PageRank Algorithm for Very Large Graphs on the Fugaku Supercomputer 1 Introduction 2 Background and Related Work 3 Parallel and Distributed Implementation 3.1 PageRank 3.2 Sparse Matrices 4 Experiments 4.1 Parameters 4.2 Results 4.3 Discussion 5 Conclusion References Ultrafast Focus Detection for Automated Microscopy 1 Introduction 2 Background 2.1 Connectomics 2.2 Scale-Space Representations 3 Multi-scale Histologic Feature Detection 4 Histologic Feature Detection as a Service 5 Evaluation 6 Related Work 7 Conclusions References Models and Metrics for Mining Meaningful Metadata 1 Introduction 2 Related Work 2.1 Metadata Extraction Systems 2.2 File Type Identification 3 Metadata Quality Determination 4 Methodology 5 Evaluation 5.1 FTI Modeling 5.2 Extractor Scheduler and Metadata Analysis 6 Conclusion References HNOP: Attack Traffic Detection Based on Hierarchical Node Hopping Features of Packets 1 Introduction 2 Related Work 3 System Scheme 3.1 Hierarchical Node Hopping Structure 3.2 G_BERT 3.3 Classifier 4 Evaluation 4.1 Data Sets 4.2 The Analysis of Hierarchical Relationship 4.3 Compared with Other Methods 5 Conclusion References TROPHY: Trust Region Optimization Using a Precision Hierarchy 1 Introduction 2 Background 2.1 Trust Region Methods 2.2 Model Function 3 Method 4 Test Problems and Implementations 4.1 CUTEst 4.2 Multiple Doppler Radar Wind Retrieval: 5 Experimental Results 6 Conclusion and Future Work References Incremental Mining of Frequent Serial Episodes Considering Multiple Occurrences 1 Introduction 2 Basic Concepts and Problem Statement 3 Incremental Algorithm for Sequential Patterns 3.1 Illustration of the Algorithm 3.2 Merging a Tree of an Itemset into a Tree of Frequent Patterns 3.3 Completion of a List of Occurrences 4 Experiments and Results 4.1 Experiments on Synthetic Data 4.2 Experiments on Smart Electrical Meter Data 5 Related Work 6 Conclusion and Future Works References MHGEE: Event Extraction via Multi-granularity Heterogeneous Graph 1 Introduction 2 Related Work 3 Approach of MHGEE Model 3.1 Input Layer 3.2 Graph Construction 3.3 Information Aggregation over MHG 3.4 Classification Layer 3.5 Biased Loss Function 4 Experiments and Results 4.1 Experiment Settings 4.2 Baselines 4.3 Overall Performance and Ablation Analysis 4.4 Effect on Event Interdependency 4.5 Case Study and Effect on Ambiguity of Triggers 5 Conclusions and Future Works References Consistency Fences for Partial Order Delivery to Reduce Latency 1 Introduction 2 Related Work 3 Consistency Fences 4 Combining Partial Order Delivery with Forward Error Correction 5 Evaluation 6 Concluding Remarks References Dynamic Classification of Bank Clients by the Predictability of Their Transactional Behavior 1 Introduction 2 Related Work 3 The Description of the Proposed Method 4 Experimental Study 5 Conclusions References Unveiling User Behavior on Summit Login Nodes as a User 1 Introduction 2 Characteristics of the Summit Login Node Data 3 System Metrics 3.1 Users Access 3.2 I/O Throughput 3.3 Computational Jobs 4 User Behavior 4.1 User Sessions 4.2 Misuse 4.3 Scientific Workflows 5 Related Work 6 Conclusion and Future Work References n-type B-N Co-doping and N Doping in Diamond from First Principles 1 Introduction 2 Calculation Methods 3 Results and Discussion 3.1 Impurity Formation Energy(Ef) 3.2 Band Structure 3.3 Ionization Energies 3.4 Electronic Structure 4 Conclusions References Stock Predictor with Graph Laplacian-Based Multi-task Learning 1 Introduction 2 Related Works 3 Methods 3.1 Problem Formulation 3.2 Federated Multi-task Stocks Predictor 3.3 Graph Laplacian Interpretation for Financial Tasks 4 Experiments 4.1 Experiment Settings 4.2 Results 5 Conclusion References Identification of MEEK-Based TOR Hidden Service Access Using the Key Packet Sequence 1 Introduction 2 Background and Related Work 2.1 Tor Hidden Service 2.2 Meek 2.3 Related Work 3 Methodology 3.1 Threat Model 3.2 Key Sequence Definition 3.3 Key Sequence Based Identification Methodology 4 Data Collection and Processing 4.1 MEEK Access Scenarios Configuration 4.2 Data Collection 4.3 Data Extraction and Processing 5 Evaluation and Discussion 5.1 Select Classification Model 5.2 Key Sequence Verification 5.3 Key Sequence Location Distribution Observation 5.4 Classification with Key Sequence 6 Conclusion References Towards a Scalable Set Similarity Join Using MapReduce and LSH 1 Introduction 2 Preliminaries 2.1 Locality Sensitive Hashing (LSH) 2.2 Distributed Histograms and Communication Templates 2.3 MRS-join 3 MRSS-join: A Scalable Set Similarity Join Algorithm Using LSH and MapReduce 4 Experiments 4.1 Performance and Results Quality 5 Conclusion References Cyberbullying Detection with Side Information: A Real-World Application of COVID-19 News Comment in Chinese Language 1 Introduction 2 Related Work 3 Problem Definition 4 The Proposed Model 4.1 Semantic Context Encoding 4.2 Digital Data Encoding 4.3 Session Prediction 4.4 Training Process 5 Experiments 5.1 JRTT: A New Benchmark Dataset 5.2 Experimental Settings 5.3 Experimental Results and Analysis 5.4 Repeatability 6 Conclusion References Designing a Training Set for Musical Instruments Identification 1 Introduction 1.1 Related Work 1.2 Contribution and Paper Structure 2 Identification of Musical Instruments 2.1 Extraction of Sound Features 2.2 Classification Method 3 Experiments 3.1 Dataset and Hardware Setup 3.2 Assessment of the Decision Confidence 3.3 Impact of Sound Characteristics 3.4 The Influence of Training Data Volume 3.5 The Impact of Data Type 4 Conclusions References Characterizing Wildfire Perimeter Polygons from QUIC-Fire 1 Introduction 2 QUIC-Fire Output Data 3 Polygon Algorithms for Wildfire Perimeters 3.1 Image Data 3.2 Quadriculation Algorithm 3.3 Iterative Minimum Distance Algorithm (IMDA) 4 Numerical Results 5 Conclusions References On a New Generalised Iteration Method in the PSO-Based Newton-Like Method 1 Introduction 2 PSO-Based Newton-Like Method 3 Iteration Processes 4 The Generalised Iteration 5 Experimental Setup 6 Discussion on the Obtained Results 6.1 Minimising the Number of Iterations 6.2 Obtaining an Aesthetic Pattern 7 Conclusions References Peridynamic Damage Model Based on Absolute Bond Elongation 1 Introduction 2 The Peridynamic Theory 3 The Peridynamic with Damage 3.1 Determination of Elastic Constants 3.2 Determination of Damage Constants 3.3 Conservation of Energy and Energy Decay 4 Numerical Examples 4.1 Example 1 4.2 Example 2 5 Conclusions References Privacy Paradox in Social Media: A System Dynamics Analysis 1 Introduction 2 Theoretical Background 2.1 Informational Privacy 2.2 Privacy Paradox 3 A System Dynamics Model of the Privacy Paradox 3.1 Problem Articulation 3.2 Dynamic Hypothesis 4 Model Development 4.1 Model Structure 4.2 Model Parameters 4.3 Model Testing and Validation 5 Simulation Results 5.1 Simulation Experiment 1 5.2 Simulation Experiment 2 5.3 Simulation Experiment 3 5.4 Simulation Experiment 4 5.5 Sensitivity Analysis 6 Concluding Discussion References GPU Power Capping for Energy-Performance Trade-Offs in Training of Deep Convolutional Neural Networks for Image Recognition 1 Introduction 2 Related Works 2.1 Power Capping for GPU Servers 2.2 Energy Aware Machine Learning 3 Testbed Systems 4 Deep CNN Benchmarks Used for Experiments 5 Experiments and Results 5.1 Power Monitoring and Controlling Methodology 5.2 Target Metrics for Bi-Objective Energy-Performance Optimization 5.3 Results Obtained for Quadro 6000 System 5.4 Results Obtained for V100 System 6 Conclusions and Future Work References Forecasting Bank Default with the Merton Model: The Case of US Banks 1 Introduction 2 Data and Methodology 2.1 Data 2.2 Hypotheses 2.3 Merton’s DD Probability and Its Extension 2.4 Hazard Model 3 Results and Discussion 3.1 Hazard Model Results 3.2 Out-of-Sample Results 4 Conclusion References .26em plus .1em minus .1emDeep Neural Sequence to Sequence Lexical Substitution for the Polish Language 1 Introduction 2 Related Work 2.1 Full Masking 2.2 Partial Masking 2.3 Cross-sentence Relationship 3 Multi-token Prediction 4 Datasets 5 Sequence-to-Sequence Substitution 6 Synthetic Dataset Generation 6.1 Single-word Replacement (SWR) 6.2 Single Word and Context Replacement (SWCR) 6.3 Substitution of a Longer Sentence Fragment (SLSF) 6.4 Wordnet-based Dataset Generation 7 Experiment Setup and Initial Results 8 Evaluation 9 Results and Future Work References Facial Mask Impact on Human Age and Gender Classification 1 Introduction 2 Idea 3 Baseline Validation 4 The Results of Age Groups and Gender Classification 4.1 Black Fabric Masks 4.2 Masks in Different Colors 4.3 Masks in Different Colors and Textures 5 The Performance Results Targeted at Each Model Separately 5.1 Deep Neural Network Model (DNN) 5.2 One Convolutional Layer Model (CNN1) 5.3 Three Convolutional Layers Model (CNN3) 5.4 Four Convolutional Layers Model (CNN4) 6 Conclusion References Simple and Efficient Acceleration of the Smallest Enclosing Ball for Large Data Sets in E2: Analysis and Comparative Results 1 Introduction 2 Proposed Preprocessing 2.1 Convex Hull Diameter Estimation 2.2 Theoretical Analysis 2.3 Implementation Notes 3 Experimental Results 4 Conclusion References Elastic Resource Allocation Based on Dynamic Perception of Operator Influence Domain in Distributed Stream Processing 1 Introduction 2 Motivation 2.1 Dynamic Correlation Between Upstream and Downstream Operators 2.2 Dynamic Influence Domain of Upstream Operators 2.3 Importance of Dynamic Perception of Operator Influence Domain 3 Design and Implementation 3.1 Overview 3.2 AOPartitioner: Adaptive Operator Partitioner 3.3 OLPredictor: Online Load Predictor 3.4 OPPlanner: Online Partition-based Operator Parallelism Planner 3.5 Scheduler 4 Experiments 4.1 Settings and Datasets 4.2 Evaluation 4.3 AOPartitioner Evaluation 4.4 OLPredictor Evaluation 4.5 OPPlanner Evaluation 5 Conclusion References PRISM: Principal Image Sections Mapping 1 Introduction 2 State-of-the-Art 3 Problem Description 4 Method 5 Experiments 6 Clustering Utility 7 Conclusions 8 Future Works References The New UPC++ DepSpawn High Performance Library for Data-Flow Computing with Hybrid Parallelism 1 Introduction 2 Related Work 3 Data-flow Computing with UPC++ DepSpawn 3.1 UPC++ 3.2 UPC++ DepSpawn 4 An Improved Runtime 4.1 Shared Memory Parallelism Migration 4.2 Distributed Memory Parallelism Migration 5 Evaluation 6 Conclusions References Author Index The four-volume set LNCS 13350, 13351, 13352, and 13353 constitutes the proceedings of the 22ndt International Conference on Computational Science, ICCS 2022, held in London, UK, in June 2022.* The total of 175 full papers and 78 short papers presented in this book set were carefully reviewed and selected from 474 submissions. 169 full and 36 short papers were accepted to the main track; 120 full and 42 short papers were accepted to the workshops/ thematic tracks. *The conference was held in a hybrid format