Engineering Applications of Neural Networks: 23rd International Conference, EAAAI/EANN 2022, Chersonissos, Crete, Greece, June 17–20, 2022, ... in Computer and Information Science, 1600)

Preface
Organization
Abstracts of Invited Talks
	Machine Learning: A Key Ubiquitous Technology in the 21st Century
	What Neuroimaging Can Tell About Human Brain Function
	Socially Interactive Artificial Intelligence: Perception, Synthesis and Learning of Human-Like Behaviors
	Responsible Conversational AI: Trusted, Safe and Bias-free
	Is Big Tech Becoming the Big Tobacco of AI?
	Contents
Bio Inspired Modeling/Novel Neural Architectures
Evaluating Acceleration Techniques for Genetic Neural Architecture Search
	1 Introduction
	2 Related Work
		2.1 Evolutionary Neural Architecture Search with Fitness Approximation
		2.2 Neural Architecture Search Without Training
		2.3 NAS-Bench-101
	3 Experiments
		3.1 Genetic Algorithm
		3.2 NAS-EA-FA V2
		3.3 Experimental Setup
	4 Results
	5 Conclusions and Future Work
	References
Generation of Orthogonality for Feature Spaces in the Bio-inspired Neural Networks
	1 Introduction
	2 Bio-inspired Neural Networks
		2.1 Background of Bio-inspired Neural Networks
	3 Orthogonal Characteristics of Asymmetric Networks
		3.1 Orthogonality of Asymmetric Network Under the Stimulus Condition
		3.2 Orthogonality Between the Asymmetric Networks Units
	4 Orthogonal Properties of Conventional Energy Model
		4.1 Orthogonality Under the Stimulus Condition
		4.2 Comparison Between the Asymmetric Network and the Energy Model
	5 Generation of Orthogonality in the 1st Layer Network
		5.1 Orthogonality in the 1st Layer in the Model
	6 Extension of Asymmetric Networks Based on Bio-inspired Neural Networks
		6.1 Extension of the Asymmetric Networks to the 2-nd and 3-rd Layers
		6.2 Generation of Orthogonality in the 2ndand 3rdLayers
		6.3 Tracking Characteristics in the Conventional Symmetric Network
	7 Conclusion
	References
Route Scheduling System for Multiple Self-driving Cars Using K-means and Bio-inspired Algorithms
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Bio-inspired Algorithms
		3.2 Hybrid Algorithms k-means-GA and k-means-ACS
		3.3 Route Scheduling System
	4 Experimental Results
	5 Conclusions
	References
SNNs Model Analyzing and Visualizing Experimentation Using RAVSim
	1 Introduction
	2 Related Work
	3 RAVSim Tool
	4 Discussion and Results
		4.1 NLIF Neuronal Model
		4.2 Learning Mechanism Using WTA Network
		4.3 Simulator Comparison
	5 Conclusion
	References
The Effectiveness of Synchronous Data-parallel Differentiable Architecture Search
	1 Introduction
	2 Background
		2.1 DARTS
		2.2 Fashion-MNIST in NAS
	3 Methodology and Experimental Setup
	4 Results
		4.1 Search Phase
		4.2 Best Model Analysis
	5 Limitations
	6 Discussion and Future Work
	References
Classification/Clustering - Machine Learning
Automatic Accent and Gender Recognition of Regional UK Speakers
	1 Introduction
	2 Related Work
		2.1 Importance of Accent/Dialect Classification
		2.2 Automatic Methods for Accent/Dialect Classification
	3 Methodology
		3.1 Data and Features Extraction
		3.2 Machine Learning Classification Models
	4 Results and Discussion
	5 Conclusions
	References
Complex Layers of Formal Neurons
	1 Introduction
	2 Linear Separability With Margin
	3 Complex Layers of Formal Neurons
	4 Perceptron Penalty Functions with Zero Threshold
	5 Dual Hyperplanes and Vertices in the Parameter Space
	6 Minimization of the Perceptron Criterion Function
	7 Selected Properties of Optimal Vertices
	8 Concluding Remarks
	References
Novel Decision Forest Building Techniques by Utilising Correlation Coefficient Methods
	1 Introduction
	2 Related Work
	3 Proposed Methods
		3.1 Splitting Criteria
		3.2 Function Features Importance
		3.3 Function growTree
	4 Experimental Setup and Results
	5 Conclusion
	References
On Forecasting Project Activity Durations with Neural Networks
	1 Introduction
	2 Related Work
	3 Problem Formulation
	4 Method
	5 Experiments
	6 Results
	7 Discussion and Conclusions
	References
On the Suitability of Neural Networks as Building Blocks for the Design of Efficient Learned Indexes
	1 Introduction
		1.1 Computer Architectures as a Motivation for Learned Data Structures
		1.2 From Motivation to Design and Implementation of Learned Data Structures: The Role of NNs
		1.3 Our Results: The Role of Neural Networks in the Design of Learned Indexes - The Atomic Case
		1.4 Organization of the Paper
	2 Learned Indexes: A Synopsis
	3 Atomic Models for Learned Indexes
		3.1 CDF Function Models Based on Analytic Solutions to Regression Problems
		3.2 CDF Function Models Based on Neural Networks
		3.3 Prediction Accuracy of an Atomic Model
	4 Experimental Methodology
		4.1 Hardware and Datasets
		4.2 Binary Search and the Corresponding Atomic Learned Indexes
	5 Experiments and Findings
		5.1 Training: GPU vs CPU
		5.2 Query: GPU only for NNs
		5.3 Query: CPU only for All Atomic Models
	6 Conclusions
	References
Supporting Patient Nutrition in Critical Care Units
	1 Introduction
	2 Problem Definition
	3 Related Work
	4 An Automation Prototype
	5 Applying Machine Learning
		5.1 Feature Extraction
		5.2 Feature Engineering
		5.3 Model Training
		5.4 Evaluation
		5.5 Results
	6 Sensitivity Analysis
	7 Discussion and Conclusions
	References
Convolutional/Deep Learning
An Accurate Convolutional Neural Networks Approach to Wound Detection for Farmed Salmon
	1 Introduction
		1.1 Drawbacks
	2 Dataset
		2.1 Datasets Creation
	3 Methodology
		3.1 Preprocessing
		3.2 Convolutional Neural Network Model
	4 Result
	5 Conclusion
	References
Autoregressive Deep Learning Models for Bridge Strain Prediction
	1 Introduction
	2 Area of Research – Literature Review
	3 Dataset
		3.1 Dataset Preprocessing
	4 Algorithmic Approaches
		4.1 Machine Learning - Deep Learning Algorithms
		4.2 Evaluation of Deep Learning Algorithms
	5 Evaluation and Experimental Results
	6 Conclusions and Future Work
	References
Ground Penetrating Radar Fourier Pre-processing for Deep Learning Tunnel Defects\' Automated Classification
	1 Introduction
	2 Monitoring Road Tunnels with GPR
	3 Two Dimensional Fourier Transform for Image Processing
	4 Methodology
		4.1 Multi-level Defect Classification
		4.2 Convolutional Neural Network: ResNet50
	5 Results and Discussion
		5.1 Confusion Matrix for Each Level
		5.2 Comparison with the Authors\' Previous Work
	6 Conclusions
	References
ITSC Fault Diagnosis in Permanent Magnet Synchronous Motor Drives Using Shallow CNNs
	1 Introduction
	2 Literature Review
	3 The Proposed Method
	4 Experiment and Results
		4.1 Approach 1 - Non-overlapping Samples
		4.2 Approach 2 - Sliding Windows
		4.3 Comparison with Model-based Approach
	5 Conclusions
	References
Synopsis of Video Files Using Neural Networks
	1 Introduction
		1.1 Current Challenges and Opportunities
		1.2 Proposed Work
	2 Related Work
	3 Technical Background
		3.1 Mixture of Gaussian Models for Background Subtraction
		3.2 Convolutional Neural Network for Visual Feature Extraction
		3.3 You Only Look Once v3 for Object Localization
		3.4 Deep Simple Online and Realtime Tracking for Object Tracking
	4 Implementation of the Synopsis Framework
		4.1 Datasets and Training Details
		4.2 Video Synopsis Algorithm
	5 Experimental Results and Discussions
		5.1 Control Datasets
		5.2 Metrics
		5.3 Results
	6 Conclusions
	References
TrojanDroid: Android Malware Detection for Trojan Discovery Using Convolutional Neural Networks
	1 Introduction
	2 Related Work
	3 Dataset
	4 Proposed Method
	5 Experimental Evaluation
		5.1 Evaluation Metrics
		5.2 Results
		5.3 Comparisons with Other Classifiers
	6 Conclusions
	References
Datamining/Learning/Autoencoders
A Methodology to Manage Structured and Semi-structured Data in Knowledge Oriented Graph
	1 Introduction
	2 Related Work
	3 Method
	4 Experiments
	5 Conclusion
	References
Discriminant Analysis on a Stream of Features
	1 Introduction
		1.1 Quadratic Discriminant Analysis
	2 Related Work
		2.1 Unrelated Work
	3 QDA Algorithm
		3.1 Covariance
		3.2 Inverse
		3.3 Determinant
		3.4 Vectorization
		3.5 Online Version
		3.6 Regularization
	4 Experiments
		4.1 Speed
		4.2 Accuracy
	5 Discussion
		5.1 Feature Selection
		5.2 Limitations
	6 Conclusion
	References
Learning Image Captioning as a Structured Transduction Task
	1 Introduction
	2 Background on Learning for Structured Data
	3 Image Captioning as Structured Transduction
		3.1 Generating Tree-Structured Image Representations
		3.2 Deep Model for Tree-Structured Captioning
	4 Empirical Analysis
		4.1 Dataset Preparation and Experimental Setup
		4.2 Results
	5 Conclusions
	References
Reconstructing Electricity Profiles in Submetering Systems Using a GRU-AE Network
	1 Introduction
	2 Related Work
	3 The Proposed Method
	4 Experiments and Results
		4.1 Submetering System and Dataset
		4.2 Experiments
		4.3 Results and Discussion
	5 Conclusions
	References
Trend and Seasonality Elimination from Relational Data
	1 Introduction
	2 Propositionalization
	3 Detrending and Deseasoning
	4 Method
	5 Experiments
	6 Results
	7 Discussion
		7.1 When the Method Works
		7.2 When the Method Fails
		7.3 Limitations of the Study
	8 Conclusion
	References
Deep Learning/Blockchain
A Blockchained Secure and Integrity-Preserved Architecture for Military Logistics Operations
	1 Introduction
	2 Proposed Architecture
		2.1 The Blockchain Architecture
		2.2 Encryption Architecture
	3 Secure and Private Transaction to Transfer Ammunition
	4 Conclusion
	References
An IoT Authentication Framework for Urban Infrastructure Security Using Blockchain and Deep Learning
	1 Introduction
	2 Literature Review
	3 Architecture of the Urban Infrastructure Security Framework (UINSE)
		3.1 The Blockchain and Deep Learning Subsystems
	4 Discussion
	5 Conclusion and Future Work
	References
Human Activity Recognition Under Partial Occlusion
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Occlusion of Skeletal Data
		3.2 Regression of Skeletal Data
	4 Experiments and Results
		4.1 Dataset
		4.2 Experimental Setup and Network Training
		4.3 Results
	5 Conclusions and Future Work
	References
Using a Deep Neural Network in a Relative Risk Model to Estimate Vaccination Protection for COVID-19
	1 Introduction
	2 Related Work
	3 Preliminaries
	4 Deep Network Model for Vaccination Waning
	5 Experiments
		5.1 Data and Model
		5.2 Results
	6 Conclusion
	References
Machine Learning for Medical Images/Genome Classification
A Gene Ontology-Driven Wide and Deep Learning Architecture for Cell-Type Classification from Single-Cell RNA-seq Data
	1 Introduction
	2 Related Works
	3 Materials and Methods
		3.1 Datasets
		3.2 Data Pre-processing
		3.3 Proposed Architecture
	4 Results
		4.1 Classifier Design
		4.2 Training and Testing Procedure
	5 Conclusions and Future Works
	References
Brain Tumour Segmentation on 3D MRI Using Attention V-Net
	1 Introduction
	2 Related Work
	3 Proposed Architecture: Attention V-Net
	4 Dataset and Training
	5 Experiments and Results
		5.1 The Evaluation Criteria
		5.2 Evaluation
	6 Conclusion and Future Work
	References
Development of an Algorithmic Model to Reduce Memory and Learning Deficits on Trisomic Mice
	1 Introduction
	2 Experiments Description
		2.1 Experimental Data and Methodology
	3 Data Preprocessing
		3.1 k-Folds Cross Validation
	4 Development of Machine Learning Models
		4.1 Models Using Proteins as Independent Variables
	5 Attributes Selection
	6 Comparative Analysis
	7 Conclusion and Discussion
	References
Semantic Segmentation of Diabetic Retinopathy Lesions, Using a UNET with Pretrained Encoder
	1 Introduction
	2 Diabetic Retinopathy Lesions
	3 DL Architecture Strategy for Semantic Segmentation
	4 Experiments
	5 Results
	6 Conclusions
	References
Reinforcement/Adversarial/Echo State Neural Networks
Echo State Networks in Data Marketplaces for Digital Content Creation
	1 Introduction
		1.1 Research Proposal
		1.2 Research Structure
	2 Research Background
	3 Echo State Networks in Data Marketplaces
		3.1 System Equations
		3.2 System Learning
		3.3 System Parameters
	4 Experimental Results
		4.1 2D Function Validation
		4.2 3D Image Validation
	5 Conclusions
	References
Efficient Approaches for Data Augmentation by Using Generative Adversarial Networks
	1 Introduction
	2 Data Augmentation
		2.1 Existing Methods of Data Generation
	3 Methodology
		3.1 Generator
		3.2 Discriminator
	4 Training GANs
	5 Dataset
	6 Experiment and Result Evaluation
		6.1 Data Analysis
		6.2 TGAN Model Setup
		6.3 Performance Analysis
	7 Conclusion and Future Work
	References
Multi-track Transfer Reinforcement Learning for Power Consumption Management of Building Multi-type Air-Conditioners
	1 Introduction
	2 Power Consumption Management System for Building Multi-type Air-Conditioners
	3 Reinforcement Learning for Building Multi-type Air-Conditioners
		3.1 Q-Learning Control
		3.2 Definition of Evaluation Function
	4 Virtual Building
		4.1 Dynamic Model of Building Multi-type Air-Conditioner
		4.2 Development of Virtual Building
	5 Multi-track Method
	6 Simulation
		6.1 Simulation Settings
		6.2 Simulation Result
	7 Discussion
	8 Conclusion
	References
Predicting Seriousness of Injury in a Traffic Accident: A New Imbalanced Dataset and Benchmark
	1 Introduction
	2 Relevant Work
	3 Data Sources
	4 Creating the Dataset
		4.1 Dataset Merging
		4.2 Dealing with Missing Values
		4.3 Missing-value Imputations Based on Domain Knowledge
		4.4 Time Processing, Feature Correlation and Feature Importance
		4.5 Imputation with MissForest
	5 Baseline Models
		5.1 A Supervised Learning Model
		5.2 A Reinforcement Learning Model
	6 Evaluation
		6.1 Supervised ANN Experiments
		6.2 Reinforcement Learning Experiments
	7 Conclusions
	References
Robotics/Autonomous Vehicles, Photonic Neural Networks
A Robust, Quantization-Aware Training Method for Photonic Neural Networks
	1 Introduction
	2 Background
	3 Proposed Method
	4 Experimental Evaluation
		4.1 Image Classification
		4.2 Forecasting Financial Time Series Analysis
	5 Conclusion
	References
Improving Binary Semantic Scene Segmentation for Robotics Applications
	1 Introduction
	2 Proposed Method
	3 Experiments
		3.1 Datasets
		3.2 Implementation Details
		3.3 Experimental Results
	4 Conclusions
	References
Online Route Scheduling for a Team of Service Robots with MOEAs and mTSP Model
	1 Introduction
	2 Related Work
	3 Multi-robot Coordination
	4 Experiments and Results
	5 Conclusion
	References
Towards Using Reinforcement Learning for Autonomous Docking of Unmanned Surface Vehicles
	1 Introduction
	2 Methods
		2.1 Environment
		2.2 Agent
		2.3 Reward Function
	3 Simulator
		3.1 Sensors
		3.2 Control
	4 Experiments and Results
	5 Conclusion and Further Work
	References
Text Classification/Natural Language
An Exploration of Semi-supervised Text Classification
	1 Introduction
	2 Methods
	3 Results
		3.1 Discussion
		3.2 Conclusion
	References
A Custom State LSTM Cell for Text Classification Tasks
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 The LSTM Cell
		3.2 Our Modified LSTM Cell
		3.3 Model
		3.4 Datasets
	4 Evaluation
		4.1 Results
		4.2 Results
	5 Limitations
	6 Conclusion
	References
Enhancements on a Pipeline Approach for Abstract Meaning Representation Parsing
	1 Introduction
	2 Related Work
	3 Approach
		3.1 Concept Identification - LSTM Based
		3.2 Concept Identification - Transformer Based
		3.3 Relation Identification
	4 Experiments and Results
		4.1 Data Analysis
		4.2 Metrics
		4.3 Evaluation of Concept Identification - LSTM Model
		4.4 Evaluation of Concept Identification - Transformer Model
		4.5 Evaluation of Relation Identification
	5 Conclusions
	References
Text Analysis of COVID-19 Tweets
	1 Introduction
		1.1 COVID-19 Vaccines
		1.2 Twitter
		1.3 Sentiment Analysis
		1.4 Paper Outline
	2 Related Work
	3 Methodology
		3.1 Tweets Hydration
		3.2 Dataset
		3.3 Model
	4 Experimental Results
		4.1 Results and Discussion
	5 Conclusion
	References
Author Index