Computational Science – ICCS 2021: 21st International Conference, Krakow, Poland, June 16–18, 2021, Proceedings, Part V (Theoretical Computer Science and General Issues)

The six-volume set LNCS 12742, 12743, 12744, 12745, 12746, and 12747 constitutes the proceedings of the 21st International Conference on Computational Science, ICCS 2021, held in Krakow, Poland, in June 2021.* The total of 260 full papers and 57 short papers presented in this book set were carefully reviewed and selected from 635 submissions. 48 full and 14 short papers were accepted to the main track from 156 submissions; 212 full and 43 short papers were accepted to the workshops/ thematic tracks from 479 submissions. The papers were organized in topical sections named: Part I: ICCS Main Track Part II: Advances in High-Performance Computational Earth Sciences: Applications and Frameworks; Applications of Computational Methods in Artificial Intelligence and Machine Learning; Artificial Intelligence and High-Performance Computing for Advanced Simulations; Biomedical and Bioinformatics Challenges for Computer Science Part III: Classifier Learning from Difficult Data; Computational Analysis of Complex Social Systems; Computational Collective Intelligence; Computational Health Part IV: Computational Methods for Emerging Problems in (dis-)Information Analysis; Computational Methods in Smart Agriculture; Computational Optimization, Modelling and Simulation; Computational Science in IoT and Smart Systems Part V: Computer Graphics, Image Processing and Artificial Intelligence; Data-Driven Computational Sciences; Machine Learning and Data Assimilation for Dynamical Systems; MeshFree Methods and Radial Basis Functions in Computational Sciences; Multiscale Modelling and Simulation Part VI: Quantum Computing Workshop; Simulations of Flow and Transport: Modeling, Algorithms and Computation; Smart Systems: Bringing Together Computer Vision, Sensor Networks and Machine Learning; Software Engineering for Computational Science; Solving Problems with Uncertainty; Teaching Computational Science; Uncertainty Quantification for Computational Models *The conference was held virtually. Preface Organization Contents – Part V Computer Graphics, Image Processing and Artificial Intelligence Factors Affecting the Sense of Scale in Immersive, Realistic Virtual Reality Space 1 Introduction 2 Related Work 3 Factors Affecting the Sense of Scale in VR 4 Evaluation 4.1 Test Tasks 5 Results and Analysis 5.1 First Session 5.2 Second Session 5.3 Experts vs Non-experts 6 Conclusion References Capsule Network Versus Convolutional Neural Network in Image Classification 1 Introduction 2 Methodology 2.1 Convolutional Neural Networks 2.2 Capsule Network 2.3 Convolutional CapsNet 3 Experiments 3.1 Datasets Description 3.2 Experiment1. Study on Impact of Capsule-Specific Elements on Performance Quality 3.3 Experiment2. Comparative Performance of CapsNet and Various CNNs Types on Augmented Test Data 3.4 Experiment3. Comparison of CapsNet and CNN Models Performance on Randomly Shattered Test Data 4 Conclusions References State-of-the-Art in 3D Face Reconstruction from a Single RGB Image 1 Introduction 2 Single Image-Based 3D Reconstruction Methods 2.1 Shape-from-Shading (SFS) Based 3D Face Reconstruction 2.2 3D Morphable Face Model (3DMM) Based 3D Face Reconstruction 2.3 Deep Learning (DL) Based 3D Face Reconstruction 3 Face Databases 4 3D Face Reconstruction Methods Based on Deep Learning 4.1 UnsupNet 4.2 PRNet 4.3 DF2Net 4.4 DFDN 4.5 Evaluations 5 Conclusions and Future Research Directions References Towards Understanding Time Varying Triangle Meshes 1 Introduction 2 Related Work 3 Algorithm Overview 4 Uniform Object Sampling 5 Initial Correspondence Estimation 6 Smoothness Energy, Smoothing 7 Sampling Energy 8 Overall Energy 8.1 Pre-smoothing 9 Results 10 Applications 11 Conclusions References Semantic Similarity Metric Learning for Sketch-Based 3D Shape Retrieval 1 Introduction 2 Related Works 2.1 Sketch-Based 3D Shape Retrieval 2.2 Teacher-Student Strategy in Metric Learning 3 Method 3.1 Network Architecture 3.2 Similarity Loss 4 Experiments 4.1 Datasets 4.2 Implementation Details 4.3 Experimental Results 5 Conclusion References ScatterPlotAnalyzer: Digitizing Images of Charts Using Tensor-Based Computational Model 1 Introduction 2 Related Work 3 Tensor Field as a Computational Model 4 Components of ScatterPlotAnalyzer 5 Experiments and Results 6 Conclusions References EEG-Based Emotion Recognition Using Convolutional Neural Networks 1 Introduction 2 Background 3 Presented Approach 3.1 Dataset 3.2 Prepossessing 3.3 Replicated Model 3.4 Improvements 4 Preliminary Results 5 Conclusions and Future Work References Improving Deep Object Detection Backbone with Feature Layers 1 Introduction 2 Background and Related Work 2.1 Object Detection Frameworks 2.2 Feature Extractors 2.3 Performance Evaluation 2.4 Datasets 2.5 Resolutions 3 Methodology 3.1 Data Preparation 3.2 Preliminaries 3.3 Backbone Model Design 4 Experiments and Results 4.1 Experiment Settings 4.2 Object Detection with Low-Resolution Images 4.3 Object Detection with Various Resolutions 5 Conclusions References Procedural Level Generation with Difficulty Level Estimation for Puzzle Games 1 Introduction 1.1 Game Rules 1.2 Piece Types and Mechanics 2 Related Work 2.1 Procedural Content Generation 2.2 Difficulty Estimation 3 Method 3.1 Procedural Level Generation 3.2 Difficulty Level Estimation 4 Results 5 Discussion 6 Conclusions and Future Work References ELSA: Euler-Lagrange Skeletal Animations - Novel and Fast Motion Model Applicable to VR/AR Devices 1 Introduction 1.1 State of the Art 1.2 Motivation 2 Materials and Methods 3 Results and Discussion 3.1 Euler-Lagrange Skeletal Animation Concept 3.2 ELSA Implementation 4 Conclusions and Acknowledgments References Composite Generalized Elliptic Curve-Based Surface Reconstruction 1 Introduction 2 Related Work 3 Generalized Ellipses (GEs) and Generalized Elliptic Curves (GECs) 4 Composite Generalized Elliptic Segments (CGESs) 5 Composite Generalized Elliptic Curve-Based Surface Reconstruction 5.1 Generalized Ellipse or Generalized Elliptic Curve-Based Surface Reconstruction 5.2 Composite Generalized Elliptic Segment-Based Surface Reconstruction 5.3 Human Body Model Reconstruction with Composite Generalized Elliptic Curves 6 Conclusions References Supporting Driver Physical State Estimation by Means of Thermal Image Processing 1 Introduction 2 Method Description 2.1 Assumptions 2.2 General Overview 2.3 Region Detectors 2.4 Eyes and Mouth State Estimation 2.5 Drowsiness Factor Estimation 3 Experiments 3.1 Dataset 3.2 Evaluation Protocol 3.3 Results 4 Conclusions References Smart Events in Behavior of Non-player Characters in Computer Games 1 Introduction 2 Related Work 2.1 Primed Agent 2.2 Smart Events 3 Method 3.1 Finite State Machine 3.2 The Proposed Solution 3.3 Experiment: The ``School'' Scenario 3.4 Experiment: The ``Shop'' Scenario 3.5 Performance Measures 4 Results and Discussion 4.1 Comparison to FSM 4.2 Improved Smart Events 5 Conclusions References Place Inference via Graph-Based Decisions on Deep Embeddings and Blur Detections 1 Introduction 2 Relevant Work 3 Algorithm and Experimental Setup 3.1 Blur Detection 3.2 Minimum Spanning Tree-Based Place Recognition 3.3 The Dataset 3.4 Algorithm 4 Experimental Results 5 Conclusions References Football Players Movement Analysis in Panning Videos 1 Introduction 2 Related Works 2.1 Field Registration 2.2 Player Detection and Reidentification 3 Method Description 3.1 Pitch Segmentation 3.2 Camera Angle Modelling 3.3 Camera Calibration 3.4 Players Detection 3.5 Players Classification 4 Experimental Results 4.1 Experimental Environment 4.2 Results 5 Limitations and Future Work 6 Conclusions References Shape Reconstruction from Point Clouds Using Closed Form Solution of a Fourth-Order Partial Differential Equation 1 Introduction 2 Related Work 3 Mathematical Model and Closed Form Solution 4 Shape Reconstruction and Error Analysis 5 Conclusion References Data-Driven Computational Sciences Addressing Missing Data in a Healthcare Dataset Using an Improved kNN Algorithm 1 Introduction 2 Related Works 3 Methods 4 Experimentation Set-up 5 Results 6 Conclusion and Future Works References Improving Wildfire Simulations by Estimation of Wildfire Wind Conditions from Fire Perimeter Measurements 1 Introduction 2 Wildfire Perimeter and Error Quantification 2.1 Uncertainty Characterization 2.2 Perimeter Interpolation 2.3 Weighted Least Squares Error 3 Wind Condition Estimation with FARSITE 3.1 Forward Simulations 3.2 Wind Speed and Wind Direction Optimization 4 Numerical Results 4.1 Maria Fire 4.2 Cave Fire 5 Conclusions References Scalable Statistical Inference of Photometric Redshift via Data Subsampling 1 Introduction 2 Photometric Redshift Estimation 3 Statistical Methodology 3.1 Partitioning the Input Space 3.2 Conditional Sampling from Partitions 3.3 Modeling via Ensembles 4 Results 5 Discussion and Conclusion References Timeseries Based Deep Hybrid Transfer Learning Frameworks: A Case Study of Electric Vehicle Energy Prediction 1 Introduction 1.1 Transfer Learning Models Review 2 Methodology 2.1 Transfer Learning 2.2 Hybrid Deep Learning Models for Transfer Learning 2.3 Baseline Convolutional Neural Networks (CNN) 2.4 Hybrid Convolutional Bidirectional Long Short-Term Memory (Conv-BiLSTM) 2.5 Hybrid Bidirectional Deep Convolutional Neural Network Long Short-Term Memory (CNN-BiLSTM) 3 Data-Sets Description 3.1 Input Processing 3.2 Data Standardisation 3.3 Evaluation Metrics 4 Results and Discussion 4.1 Training with Full Data Complement 4.2 Comparing Transfer Learning and Non-transfer Learning-Based Methods 4.3 Comparing Transfer Learning-Based Models 4.4 Summary of Findings 5 Conclusions and Future Works References Hybrid Machine Learning for Time-Series Energy Data for Enhancing Energy Efficiency in Buildings 1 Introduction 2 Hybrid Machine Learning Model 2.1 SARIMA Model 2.2 Support Vector Regression Model 2.3 Firefly – Based Optimization Algorithm 2.4 Proposed Hybrid Machine Learning Model 3 Dataset and Model Evaluation Results 3.1 Dataset 3.2 Model Evaluation Results 4 Conclusions References I-80 Closures: An Autonomous Machine Learning Approach 1 Introduction 2 Prior Work 3 Methods 4 Results 5 Conclusions and Future Work References Energy Consumption Prediction for Multi-functional Buildings Using Convolutional Bidirectional Recurrent Neural Networks 1 Introduction 1.1 Deep Learning Models Review 2 Proposed Hybrid Convolutional Bidirectional Long Short-Term Memory (Conv-BiLSTM) 3 Forecasting Methods 3.1 Baseline Convolutional Neural Networks (CNN) 3.2 Baseline Long Short-Term Memory (LSTM) 3.3 Bidirectional LSTM (BiLSTM) 4 Datasets Description 4.1 Input Processing 4.2 Data Standardisation 4.3 Evaluation Metrics 5 Results and Discussion 5.1 Implementation Details 5.2 Prediction Using 5-Min Resolution 5.3 Prediction Using 15-Min Resolution 5.4 Prediction Using Hourly Resolution 5.5 Model Parameters 6 Conclusion and Future Works References Machine Learning and Data Assimilation for Dynamical Systems Deep Learning for Solar Irradiance Nowcasting: A Comparison of a Recurrent Neural Network and Two Traditional Methods 1 Introduction 2 Methods 2.1 Deep Neural Networks in the Nowcasting Domain 2.2 Sequence-to-Sequence Model 2.3 Convolutional Gated Recurrent Unit 2.4 Model Extension 2.5 Optical Flow Baseline 2.6 The WRF Baseline Model 2.7 Irradiance Data 3 Experimental Setup 3.1 Training Set 3.2 Forecast Evaluation 4 Results 4.1 The ConvGRU Experiments 4.2 Comparison to the Baseline Models 4.3 Weather Dependent Analysis 5 Conclusion References Automatic-differentiated Physics-Informed Echo State Network (API-ESN) 1 Introduction 2 Automatic-Differentiated Physics-Informed Echo State Network (API-ESN) 3 Reconstruction of Hidden States in a Chaotic System 4 Conclusions and Future Directions References A Machine Learning Method for Parameter Estimation and Sensitivity Analysis 1 Introduction 2 Methods 3 Simple Benchmark Model Example 3.1 Decision Tree Sensitivity Analysis Example: An HIV Infection Model 4 Discussion and Conclusion References Auto-Encoded Reservoir Computing for Turbulence Learning 1 Introduction 2 Turbulent Flow 3 Auto-Encoded Reservoir Computing 4 Results 5 Conclusions and Future Directions References Low-Dimensional Decompositions for Nonlinear Finite Impulse Response Modeling 1 Introduction 2 Problem Statement 3 Separation Algorithm 4 Short-Term Memory Separation Searching 5 Numerical Experiments 6 Conclusions and Future Work References Latent GAN: Using a Latent Space-Based GAN for Rapid Forecasting of CFD Models 1 Introduction 2 Latent GAN 3 Models 3.1 Algorithms 3.2 Model Architectures 4 Testing and Evaluation on a Real Test Case 5 Summary and Future Work References Data Assimilation in the Latent Space of a Convolutional Autoencoder 1 Introduction and Motivation 1.1 Related Works and Contribution of the Present Work 2 Kalman Filter 3 Latent Assimilation 4 Experimental Results 4.1 Dimensionality Reduction 4.2 Surrogate Model 4.3 Data Assimilation 4.4 Physical Space 5 Conclusion and Future Works References Higher-Order Hierarchical Spectral Clustering for Multidimensional Data 1 Introduction 2 Clustering Multidimensional Data 2.1 Tensor Decomposition 2.2 DBHT Clustering Algorithm 2.3 Higher-Order Hierarchical Spectral Clustering (HHSC) 3 FAO Trade Network 3.1 Data 3.2 Step-by-Step HHSC Methodology 3.3 Application of HHSC to FAO Data 4 Conclusions A Time Complexity B Data Filtering C Clustering with Growth Related Size of the Nodes References Towards Data-Driven Simulation Models for Building Energy Management 1 Introduction 2 Background 3 Proposal References Data Assimilation Using Heteroscedastic Bayesian Neural Network Ensembles for Reduced-Order Flame Models 1 Introduction 1.1 Thermoacoustics 1.2 Bayesian Deep Learning 2 Methods 2.1 Bunsen Flame Experiment 2.2 Flame Edge Model 2.3 Forced Cycle Library 2.4 Inference Using Heteroscedastic Bayesian Neural Network Ensembles 2.5 Inference Using the Ensemble Kalman Filter 3 Results and Discussion 4 Conclusions A Supplementary material: Hyperparameter settings References A GPU Algorithm for Outliers Detection in TESS Light Curves 1 Introduction 2 Detection of Transits in TESS Light Curves 3 Mathematical Background 4 GPU-Parallel Algorithm 5 Results 6 Conclusion References Data-Driven Deep Learning Emulators for Geophysical Forecasting 1 Introduction 2 Proper Orthogonal Decomposition 3 Models 3.1 Forecasting Neural Network Models 3.2 Error Correction Model 4 Dataset and Numerical Results 5 Conclusion References NVIDIA SimNetTM: An AI-Accelerated Multi-Physics Simulation Framework 1 Introduction 2 Neural Network Solvers 3 SimNet Overview 3.1 Geometry Modules 3.2 PDE Module 3.3 Network Architectures 4 Use Cases 4.1 Turbulent and Multi-physics Simulations 4.2 Blood Flow in an Intracranial Aneurysm 4.3 Design Optimization for Multi-physics Industrial Systems 4.4 Inverse Problems 5 Performance Upgrades and Multi-GPU Training 6 Conclusion References MeshFree Methods and Radial Basis Functions in Computational Sciences Analysis of Vortex Induced Vibration of a Thermowell by High Fidelity FSI Numerical Analysis Based on RBF Structural Modes Embedding 1 Introduction 2 Theoretical Background 2.1 Unsteady FSI Using Modal Superposition 2.2 RBF Mesh Morphing 2.3 Modal FSI Implementation 3 Experimental Investigation 4 Numerical Analysis 4.1 Modal Analysis 4.2 RBF Solutions Setup 4.3 CFD and FSI Setups 4.4 Damping Ratio 4.5 FSI Analysis Results 5 Conclusions References Automatic Optimization Method Based on Mesh Morphing Surface Sculpting Driven by Biological Growth Method: An Application to the Coiled Spring Section Shape 1 Introduction 2 Recall on the Theoretical Background 2.1 RBF Based Mesh Morphing 2.2 Biological Growth Method 2.3 Parameter-Less Based Optimisation 2.4 Parameter-Based Shape Optimisation 3 Coiled Springs Background 3.1 Equivalent Circular Section 3.2 Numerical Model of the Coiled Spring 4 Results 4.1 Sculpting Inner Surface 4.2 Sculpting Outer Surface 4.3 Optimization Method Comparison 5 Conclusions References Multiscale Modelling and Simulation Verification, Validation and Uncertainty Quantification of Large-Scale Applications with QCG-PilotJob 1 Introduction 2 Related Work 3 Objectives 4 Use Cases 5 Performance Evaluation 6 Summary and Future Work References Towards a Coupled Migration and Weather Simulation: South Sudan Conflict 1 Introduction 2 Flee: A Multiscale Approach 3 Coupling Approaches 3.1 File Coupling 3.2 MUSCLE3 Coupling 4 South Sudan Multiscale Simulation 4.1 Weather Coupling in South Sudan Microscale Model 5 Results and Discussion 6 Conclusion References Evaluating WRF-BEP/BEM Performance: On the Way to Analyze Urban Air Quality at High Resolution Using WRF-Chem+BEP/BEM 1 Introduction 2 Data and Methods 2.1 Modelling System 2.2 Experimental Study Case 3 Results 3.1 Quality Results 3.2 Scalability Results 4 Conclusions and Future Steps References Pathology Dynamics in Healthy-Toxic Protein Interaction and the Multiscale Analysis of Neurodegenerative Diseases 1 Introduction 2 Mathematical Model 3 Temporal Dynamics 3.1 Stationary Points 3.2 Linear Stability Analysis 4 Wave Propagation 5 Network Mathematical Model 6 Results and Discussion 7 Conclusion References A Semi-implicit Backward Differentiation ADI Method for Solving Monodomain Model 1 Introduction 2 Derivation of SBDF-ADI Method 2.1 Two-Dimensional SBDF Method 2.2 Two-Dimensional SBDF-ADI Method 3 Numerical Results 4 Conclusion References A Deep Learning Approach for Polycrystalline Microstructure-Statistical Property Prediction 1 Introduction 2 Statistical Scatter in the Homogenized Polysilicon Properties at the Mesoscale 3 Methodology 3.1 Input Data Generation and Pre-processing 3.2 Proposed Models and Implementation 3.3 ResNet vs DenseNet: Addition vs Concatenation of Features 4 Results 4.1 Generalization Capability of Trained Models 5 Conclusions References MsFEM Upscaling for the Coupled Thermo-Mechanical Problem 1 Introduction 1.1 Asphalt Concrete 1.2 Multiscale AC Analyses 2 Problem Formulation 3 Multiscale Finite Element Method 4 Numerical Results 4.1 The Thermal Part 4.2 The Mechanical Part 5 Concluding Remarks References MaMiCo: Non-Local Means Filteringpg with Flexible Data-Flow for Couplingpg MD and CFD 1 Introduction 2 Coupled Simulation 3 Non-Local Means Algorithm 4 Implementation and Software Design 4.1 Supported Types of Filters 4.2 Non-Local Means: Optimized Implementation 4.3 NLM Test: Nano-Pattern (Sound Wave) Extraction 5 Simulation Results 5.1 Vortex Street Filtering Results 6 Conclusions References Author Index The six-volume set LNCS 12742, 12743, 12744, 12745, 12746, and 12747 constitutes the proceedings of the 21st International Conference on Computational Science, ICCS 2021, held in Krakow, Poland, in June 2021.* The total of 260 full papers and 57 short papers presented in this book set were carefully reviewed and selected from 635 submissions. 48 full and 14 short papers were accepted to the main track from 156 submissions; 212 full and 43 short papers were accepted to the workshops/ thematic tracks from 479 submissions. The papers were organized in topical sections named: Part I: ICCS Main Track Part II: Advances in High-Performance Computational Earth Sciences: Applications and Frameworks; Applications of Computational Methods in Artificial Intelligence and Machine Learning; Artificial Intelligence and High-Performance Computing for Advanced Simulations; Biomedical and Bioinformatics Challenges for Computer Science Part III: Classifier Learning from Difficult Data; Computational Analysis of Complex Social Systems; Computational Collective Intelligence; Computational Health Part IV: Computational Methods for Emerging Problems in (dis- )Information Analysis; Computational Methods in Smart Agriculture; Computational Optimization, Modelling and Simulation; Computational Science in IoT and Smart Systems Part V: Computer Graphics, Image Processing and Artificial Intelligence; Data-Driven Computational Sciences; Machine Learning and Data Assimilation for Dynamical Systems; MeshFree Methods and Radial Basis Functions in Computational Sciences; Multiscale Modelling and Simulation Part VI: Quantum Computing Workshop; Simulations of Flow and Transport: Modeling, Algorithms and Computation; Smart Systems: Bringing Together Computer Vision, Sensor Networks and Machine Learning; Software Engineering for Computational Science; Solving Problems with Uncertainty; Teaching Computational Science; Uncertainty Quantification for Computational Models *The conference was held virtually