Proceedings of the 22nd Engineering Applications of Neural Networks Conference: EANN 2021 (Proceedings of the International Neural Networks Society, 3)

This book contains the proceedings of the 22nd EANN “Engineering Applications of Neural Networks” 2021 that comprise of research papers on both theoretical foundations and cutting-edge applications of artificial intelligence. Based on the discussed research areas, emphasis is given in advances of machine learning (ML) focusing on the following algorithms-approaches: Augmented ML, autoencoders, adversarial neural networks, blockchain-adaptive methods, convolutional neural networks, deep learning, ensemble methods, learning-federated learning, neural networks, recurrent – long short-term memory. The application domains are related to: Anomaly detection, bio-medical AI, cyber-security, data fusion, e-learning, emotion recognition, environment, hyperspectral imaging, fraud detection, image analysis, inverse kinematics, machine vision, natural language, recommendation systems, robotics, sentiment analysis, simulation, stock market prediction. 22nd EANN/(EAAAI New Name) 2021 Preface Organization General Co-chairs Program Co-chairs Honorary Co-chairs Liaison Co-chairs Advisory Co-chairs Workshops’ Co-chairs Publication and Publicity Chair Special Sessions/Tutorials Co-chairs Steering Committee Co-chairs Program Committee Keynote Lectures Is “Big Tech” Becoming the “Big Tobacco” of Artificial Intelligence?” Biography: Machine Learning: A Key Ubiquitous Technology in the Twenty-First Century Short Bio: Human-Centered Computer Vision: Core Components and Applications Short Bio: Unveiling Recurrent Neural Networks—What Do They Actually Learn and How? Short Bio: Deep Learning and Kernel Machines Short Bio: How Can Artificial Intelligence Efficiently Support Sustainable Development? Short Bio: Backpropagation Free Deep Learning Short Bio: Backpropagation Free Deep Learning Short Bio: Workshops Tutorials Modern Methods and Tools for Human Biosignal Analysis Short Bio: Anomaly Detection in Images Short Bio: Utilizing Field-Programmable Gate Array (FPGA) for AI Acceleration Without Noticing it! Short Bio: Contents Adversarial Neural Networks - Classification Automatic Facial Expression Neutralisation Using Generative Adversarial Network 1 Introduction 2 Related Works 3 Proposed Method 3.1 Notation 3.2 Training Strategy 3.3 Loss Functions 3.4 Network Architecture 4 Experiments 5 Conclusion References Creating Ensembles of Generative Adversarial Network Discriminators for One-Class Classification 1 Introduction 2 Related Work 3 A Novel Approach to One-Class Classification Using GANs 3.1 Motivation 4 The GANOCC Algorithm 5 Experiments 5.1 Evaluation Method 5.2 Results on the MNIST Dataset 5.3 Results on the CIFAR-10 Dataset 6 Discussion 7 Conclusion and Future Work References Anomaly Detection A Hybrid Deep Learning Ensemble for Cyber Intrusion Detection 1 Introduction 1.1 Literature Review 2 Dataset 2.1 Dataset Description 2.2 Dataset Pre-processing 3 The Hybrid Ensemble COREM Model 3.1 Architecture of the Hybrid COREM Model 3.2 Evaluation of the Proposed Model 4 Experimental Results 5 Conclusions and Future Work References Anomaly Detection by Robust Feature Reconstruction 1 Introduction 2 Methodology 2.1 Problem Definition 2.2 Autoencoder 2.3 Loss Functions 2.4 Proposed Loss Function 3 Experiments 3.1 Data 3.2 Implementation 3.3 Performance Evaluation 3.4 Results 4 Conclusions References Classification - Medical Applications Deep Learning of Brain Asymmetry Images and Transfer Learning for Early Diagnosis of Dementia 1 Introduction 2 Related Work 3 MRI Data Repositories 4 Proposed Approach 4.1 MRI Image Analysis and Segmentation of Brain Asymmetry 4.2 Transfer Learning Architecture 5 Experimental Study 5.1 Experiment 1: ADNI Repository 5.2 Experiment 2: OASIS Database 6 Discussion and Conclusions References Deep Learning Topology–Preserving EEG–Based Images for Autism Detection in Infants 1 Introduction 2 EEG Data Collection in Infants 3 Framework of Proposed Approach 3.1 Transforming EEG Signals into Topology-Preserving Images 3.2 Deep Learning Considerations 4 Experiments and Results 5 Conclusions References Improving the Diagnosis of Breast Cancer by Combining Visual and Semantic Feature Descriptors 1 Introduction 2 The Proposed CAD System 2.1 The Shapelet Transform 2.2 Angular Radial Transform 2.3 Haralick Features 2.4 Semantic Feature Representation of Mammographic Cases 2.5 Feature Extraction Using Quaternionic Representation 2.6 The RBF-NN Classifier 3 Experimental Results 3.1 The Data Set 3.2 The RBF-NN Classifier 3.3 Experimental Results 4 Conclusions References Liver Cancer Trait Detection and Classification Through Machine Learning on Smart Mobile Devices 1 Introduction 2 Related Work 3 Proposed Concept 3.1 Dataset: Data Collection 3.2 Evaluation Metrics 3.3 Methodology and Experimental Steps 4 Results and Comparisons 5 Conclusions References Using WOA with Feed Forward Neural Network in Prediction of Subcutaneous Glucose Concentration for Type-1 Diabetic Patients 1 Introduction 2 Literature Survey 3 Using Whale Optimization Algorithm 4 Description of FFNN-WOA Training Model 4.1 Subjects and Dataset 4.2 Evaluation Metrics 5 Effect of Adding WOA to FFNN in Prediction Performance 6 Algorithm Comparison 7 Conclusion References Deep Learning - Convolutional Neural Networks Modeling A Novel CNN-LSTM Hybrid Architecture for the Recognition of Human Activities 1 Introduction 2 Related Work 2.1 Visual Skeletal Representations 2.2 Multimodal Methods 3 Methodology 3.1 Input Pre-processing 3.2 Skeletal Information 3.3 Activity Images 3.4 Network Architecture 4 Experiments 4.1 Datasets 4.2 Implementation and Network Training Details 4.3 Experimental Results and Analysis 5 Conclusion References An Artificial Intelligence System for Endotracheal Intubation Confirmation 1 Introduction 2 Materials and Methods 2.1 Instruments 2.2 Algorithm 2.3 Dataset 2.4 Data Augmentation 3 Results 4 Discussion References Data Fusion for Deep Learning on Transport Mode Detection: A Case Study 1 Introduction 2 Related Works 2.1 Transport Mode Detection 2.2 Data Fusion Modes in Deep Learning 3 Baseline 3.1 A Network Using a Single Modality 3.2 Training Protocol 4 Experiments 4.1 Sensor Choice 4.2 Preprocessing 4.3 Fusion Modes 5 Results 5.1 Evaluation of Unimodal Sensors 5.2 Evaluation of Preprocessing Methods 5.3 A Benchmark of Fusion Modes 5.4 Evaluations on the Test Set 6 Conclusion References Deep Learning for Water Quality Classification in Water Distribution Networks 1 Introduction 2 Related Work 2.1 Dataset 2.2 Contamination Scenario 3 Methodology 3.1 Data Exploration 3.2 Data Representation Concerning Water Conditions (Data Labeling) 4 Modeling Implementation 4.1 Support Vector Machine (SVM) 4.2 Artificial Neural Network (ANN) 5 Results 5.1 SVM Performance 5.2 ANN Performance 6 Conclusion References Deep Learning Modeling of Groundwater Pollution Sources 1 Introduction 1.1 Theoretical Problem Definition 1.2 Research Goals 2 Hydraulic Simulation 2.1 Flow Field and Mass Transport Simulation 2.2 Building of Datasets 3 Machine/Deep Learning Implementation - Results 3.1 Machine Learning (Classification) - Random Forests, Multi-layer Perceptron 3.2 Deep Learning (Computer Vision) - Convolutional Neural Networks 4 Conclusions References Deep Neural Networks for Indoor Geomagnetic Field Fingerprinting with Regression Approach 1 Introduction 2 Related Work 3 Dataset and Preprocessing 4 Proposed DNN-Based Fingerprinting Systems 4.1 Input and Output of the Systems 4.2 Deep Neural Networks 4.3 Model Selection 5 Testing Results 6 Conclusion and Future Work References Event-Detection Deep Neural Network for OTDR Trace Analysis 1 Introduction 2 Related Work 2.1 Object Detection Networks for Images 2.2 Deep Learning for Time Series 2.3 Deep Learning for OTDR Trace Analysis 3 Problem Formulation 4 Deep Detection Network for OTDR Traces 4.1 Architecture of the Proposed Model 4.2 Training 5 Experiments 5.1 Dataset of OTDR Traces 5.2 Experimental Setup 5.3 Event Classification Performance 5.4 Event Detection Performance Metric 5.5 Event Detection Performance 6 Conclusion and Future Work References Exploring the Limits of Vanilla CNN Architectures for Fine-Grained Vision-Based Vehicle Classification 1 Introduction 2 Approach 2.1 Overview 2.2 Data Acquisition 2.3 Classification Approach 3 Results 4 Conclusion References fNIRS–Based BCI Using Deep Neural Network with an Application to Deduce the Driving Mode Based on the Driver's Mental State 1 Introduction 2 Data Capture Experiments 3 Recognition Experiments 4 Conclusions References Image Pre-processing and Segmentation for Real-Time Subsea Corrosion Inspection 1 Introduction 2 Related Work 2.1 Pre-processing 2.2 Real-Time Detection 2.3 Instance Segmentation 3 Methodology 4 Results 5 Conclusion and Future Work References Squeeze-and-Threshold Based Quantization for Low-Precision Neural Networks 1 Introduction 2 Related Work 3 Quantization Methodology 3.1 Activation Quantization: Squeeze-and-Threshold with Momentum 3.2 Weight Quantization 4 Experiments 4.1 Datasets: ImageNet 4.2 Other Datasets: Tiny ImageNet and CIFAR-10 4.3 ST Block and Activation Threshold 5 Conclusion References Toward an Augmented and Explainable Machine Learning Approach for Classification of Defective Nanomaterial Patches 1 Introduction 2 Electrospinning Process 2.1 Electrospinning Dataset Description 3 Methodology 3.1 Data Augmentation 3.2 Convolution Neural Network Architecture 3.3 Explainability of the CNN Model via xAI: Occlusion Sensitivity Analysis 3.4 Performance Metrics 4 Results 4.1 Classification Performance 4.2 Explainable Predictions of SEM Images 4.3 Comparison with the State-of-the-art 5 Conclusion 6 Abbreviations References Face Recognition- Deep Learning - Machine Vision Addressing Computer Vision Challenges Using an Active Learning Framework 1 Introduction 2 State-of-the-Art Review 3 Data Set Construction 3.1 Video Annotation Tool 3.2 Face Recognition Data 3.3 Object Detection Data 4 The Active Learning Framework 4.1 Face Recognition System 4.2 Object Detection System 4.3 The Integrated System 5 Future Challenges References Efficient Realistic Data Generation Framework Leveraging Deep Learning-Based Human Digitization 1 Introduction 2 Related Work 2.1 Synthetic Data Generation for Computer Vision Methods 2.2 3D Human Model Generation from Natural Images 2.3 Deep Learning for Human-Centric Perception 3 Proposed Method 3.1 3D Human Model Generation and Skeleton Extraction 3.2 Data Generation Through Synthesis of Real Background Images and 3D Human Models 4 Experimental Evaluation 4.1 Dataset Generation 4.2 Experimental Evaluation 5 Conclusion References Face Detection with YOLO on Edge 1 Introduction 2 Literature 3 Method 4 Experiment 4.1 Dataset 4.2 Models 5 Results and Discussion 6 Conclusion References Face Spoof Detection: An Experimental Framework 1 Introduction 2 Related Works 3 Dataset 4 Methods 4.1 Local Binary Patterns (LBP) 4.2 CNN Models 5 Results and Discussions 6 Conclusion and Future Directions References Fuzzy Logic Modleing A Fuzzy Approach to Identity Resolution 1 Introduction 2 Literature Review 3 Problem Definition for Identity Resolution 4 The Proposed Fuzzy Approach to Identity Resolution 4.1 Policing Dataset 4.2 Data Pre-processing 4.3 String Matching and Aggregated Score 4.4 Searching and Matching Criteria 4.5 Clustering, Segmentation, and Graph Analysis 5 Experimental Evaluation 5.1 Target Entity 5.2 Searching Results 5.3 Clustering, Segmentation and Graph Analysis Results 5.4 Results Summary 6 Conclusion References Early Prediction of COVID-19 Onset by Fuzzy-Neuro Inference 1 Introduction 2 Prediction Paradigm 3 Dataset 4 EFuNN Training and Test 5 Conclusion and Future Work References Intelligent Search - Smart Energy Grids Search Problems in Contemporary Power Girds 1 Introduction 2 Modelling of Power Grids 3 Answer Set Programming in a Nutshell 3.1 Syntax 3.2 Semantics 4 Real-World Problems 4.1 Load Islanding 4.2 Load Conditions After Grid Failures 4.3 Optimal Network Topology 4.4 Critical Transmission Lines 4.5 Computation Times 5 Conclusion References VIII Learning - Blockchain Blockchained Adaptive Federated Auto MetaLearning BigData and DevOps CyberSecurity Architecture in Industry 4.0 1 Introduction 2 Motivation 3 Literature Review 4 The Proposed IΙoΤ Framework 4.1 Collector 4.2 Analyzer 4.3 Neural Search Module 4.4 Messenger 4.5 Streamer 4.6 Crypto Module 4.7 Federating Module 4.8 Blockchain Module 4.9 Elastic Stack Module 4.10 Privacy Module 5 Conclusion References Incentivizing Participation to Distributed Neural Network Training 1 Introduction 1.1 Synopsis of Architecture 1.2 Our Previous Work 2 Distributed Ledger Technology 2.1 Blockchain Fundamentals 2.2 Smart Contracts 3 Proposed Architecture 3.1 Incentivization Algorithm 3.2 Distributed Proof of Identification 4 Conducted Experiments 5 Conclusions and Future Work References Machine Learning - Classification -Simulation Contaminated Soil Detection: A Proposal Using Machine Learning and Hyperspectral Imaging 1 Introduction 2 Methods 2.1 Hyperspectral Dataset 2.2 Architectures 3 Results 3.1 Investigation of the Number of Pixels per Sample 3.2 Filter Investigation 3.3 Architecture Performance 4 Conclusion References Drilling Operations Classification Utilizing Data Fusion and Machine Learning Techniques 1 Introduction 2 Data Acquisition Process 2.1 Drilling Operations Measurements 3 Data Fusion Analysis 4 Experimental Method 4.1 Statistical Model 4.2 Classification Model 5 Results 6 Discussion 7 Conclusion References Impact of Classifiers to Drift Detection Method: A Comparison 1 Introduction 2 Related Work 3 Simulation Setup 3.1 Classifiers 3.2 Dataset 3.3 Drift Detection Method 3.4 Evaluation and Metrics 4 Simulation Results 5 Conclusions References Inverse Kinematics via a Network Ensemble and Learning Method 1 Introduction 2 Related Works and Background 3 Methodology 3.1 Kinematics Modeling 3.2 Network Ensemble 3.3 Learning Process and Dataset Extraction 4 Results and Discussion 5 Conclusion References Real-Time Multimodal Emotion Classification System in E-Learning Context 1 Introduction 2 Preliminaries 2.1 Learning from Multi-modal Data Streams 2.2 Emotion Representation 3 Materials and Methods 3.1 Data Set Description 3.2 Feature Extraction 3.3 Feed-Forward Neural Network (FFNN) 3.4 Incremental Stochastic Gradient Descent (ISGD) 3.5 Decision Level Fusion 3.6 Experimental Study 3.7 Experimental Setup 4 Results, Evaluation and Discussion 5 Conclusions References Repeatable Functionalities in Complex Layers of Formal Neurons 1 Introduction 2 Linear Separability of Learning Sets 3 Perceptron Criterion Function with Regularization 4 Dual Hyperplanes and Vertices in the Parameter Space 5 Optimal Vertices in the Parameter Space 6 Constrained Minimization of the Regularized Function 7 Separating Hyperplanes with the Largest Margins 8 Complex Layers of Formal Neurons 9 Experimental Results 10 Concluding Remarks References Using Artificial Neural Network to Provide Realistic Lifting Capacity in the Mobile Crane Simulation 1 Introduction 2 Related Work 2.1 Realism in Crane Simulations 2.2 Using Artificial Neural Network to Support Realism in Virtual Reality 3 Mobile Crane Simulation 3.1 Initial Simulation 3.2 Modified Simulation 4 Experimental Procedure 5 Results 6 Discussion 7 Limitations and Future Work 8 Conclusion References Recommendation systems A Multi-modal Audience Analysis System for Predicting Popularity of Online Videos 1 Introduction 2 Related Work 2.1 Methods for Audience Analysis 2.2 Video Views Prediction 3 Proposed Framework 3.1 Overview of the System 3.2 Image Processing Techniques for Audience Analysis 3.3 Audio Event Analysis 3.4 Popularity Prediction 4 Experimental Evaluation 4.1 Datasets 4.2 Experiments 5 Conclusion References Recommender Systems Algorithm Selection Using Machine Learning 1 Introduction 2 Related Work 3 Proposed Method 4 Experimental Evaluation 5 Conclusions References Sentiment Analysis - Natural Language -Financial Domain Do Weibo Platform Experts Perform Better at Predicting Stock Market? 1 Introduction 2 Related Work 2.1 Sentiment Analysis Using Neural Networks 2.2 Machine Learning in Stock Market Prediction 3 Stock Market Prediction System 3.1 Data Collection Module 3.2 Data Cleaning Module 3.3 Sentiment Analysis Module 3.4 LSTM Stock Prediction Module 3.5 Implementation Details 4 Methodology 4.1 User Grouping Method 4.2 Daily Sentiment Analysis Method 4.3 Time Window Calculation Method 4.4 Stock Prediction Model Evaluation Metrics 5 Results 6 Threats to Validity 7 Conclusions and Future Work References Predicting Stock Price Movement Using Financial News Sentiment 1 Introduction 2 Related Work 3 Methodology 3.1 Feature Selection 3.2 News Data Processing 4 Experiments and Evaluation 4.1 Experiment Settings 4.2 Results and Discussion 4.3 Evaluation 5 Conclusion and Future Work References Author Index