Neural Information Processing: 29th International Conference, ICONIP 2022, Virtual Event, November 22–26, 2022, Proceedings, Part VII

Preface
Organization
Contents – Part VII
Applications II
An Interpretable Multi-target Regression Method for Hierarchical Load Forecasting
	1 Introduction
	2 Methods
		2.1 Data and Problem Definition
		2.2 Correlation Coefficient Calculating
		2.3 Dynamically Adjusted LightGBM Regressor Chain
		2.4 SHapley Additive exPlanation (SHAP)
	3 Experiment Framework
	4 Results and Discussions
		4.1 Correlation Matrix
		4.2 Model Interpretability
	5 Conclusion and Future Work
	References
Automating Patient-Level Lung Cancer Diagnosis in Different Data Regimes
	1 Introduction
	2 Related Works
	3 Method
		3.1 DeepLung
		3.2 MilLung
		3.3 AutoLung
	4 Experimental Setup
		4.1 Dataset
		4.2 Preprocessing and Augmentation
		4.3 Hyperparameters of the Models
	5 Results
		5.1 Malignancy Classification or Regression?
		5.2 Comparison with Experienced Doctors
		5.3 Explainability
	6 Conclusions
	References
Multi-level 3DCNN with Min-Max Ranking Loss for Weakly-Supervised Video Anomaly Detection
	1 Introduction
	2 Proposed Method
		2.1 Divide Video into Segments
		2.2 Multi-level 3DCNN Feature Extraction
		2.3 Temporal Modeling
		2.4 Anomaly Detection
		2.5 Network Optimization with Min-Max Ranking Loss
	3 Experiments
		3.1 Dataset Description and Evaluation Metric
		3.2 Implementation Details
		3.3 Experimental Analysis
	4 Conclusion
	References
Automatically Generating Storylines from Microblogging Platforms
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 Event Extraction
		3.2 Story Branch Construction
		3.3 Storyline Generation
	4 Experiment
		4.1 Performance of Event Extraction
		4.2 Performance of Story Branch Construction
		4.3 Performance of Storyline Generation
	5 Conclusion
	References
Improving Document Image Understanding with Reinforcement Finetuning
	1 Introduction
	2 Related Work
	3 Background
	4 Method
		4.1 Problem Formulation for Reinforcement Finetuning
		4.2 Reward Functions
		4.3 Transfer Learning with Ground-Truth Labels
		4.4 Expert Feedback as Reward
		4.5 Training Algorithm
	5 Settings and Evaluation Metrics
	6 Experimental Results
		6.1 Transfer Learning with Limited Data
		6.2 Learning with Expert Feedback
	7 Conclusion
	References
MSK-Net: Multi-source Knowledge Base Enhanced Networks for Script Event Prediction
	1 Introduction
	2 Related Work
	3 Problem Definition
	4 Methodology
		4.1 Question Encoder
		4.2 Knowledge Searcher
		4.3 Knowledge Encoder
		4.4 Result Predictor
		4.5 Training Object
	5 Experiments
		5.1 Dataset
		5.2 Baselines
		5.3 Experimental Settings
	6 Results and Analysis
		6.1 Overall Results
		6.2 Comparative Experiments
	7 Conclusion and Future Work
	References
Vision Transformer-Based Federated Learning for COVID-19 Detection Using Chest X-Ray
	1 Introduction
	2 Related Work
	3 Chest X-Ray Findings in COVID-19
	4 Material and Methods
		4.1 Overview of ViT Architecture
		4.2 The Proposed Federated Learning Overview
	5 Experimental Result and Analysis
		5.1 Dataset Preparation
		5.2 Results
		5.3 Results on IID Data
		5.4 Results on Non-IID Data
		5.5 Results on Unbalanced Data
		5.6 Comparisons with Existing Methods
	6 Conclusion and Future Work
	References
HYCEDIS: HYbrid Confidence Engine for Deep Document Intelligence System
	1 Introduction
	2 Background
	3 Methodology
		3.1 Multi-modal Conformal Predictor (MCP)
		3.2 Variational Cluster-Oriented Anomaly Detector (VCAD)
		3.3 Hybrid Confidence Estimation
	4 Experiments
		4.1 Datasets and Evaluation Metrics
		4.2 Experimental Baselines
		4.3 Benchmarking Results
	5 Conclusion
	References
Multi-level Network Based on Text Attention and Pose-Guided for Person Re-ID
	1 Introduction
	2 Method
		2.1 Image Feature Extraction
		2.2 Text Feature Extraction
		2.3 Cross-Modal Projection Matching
	3 Experimental
		3.1 Experimental Setup
		3.2 Experimental Details
		3.3 Comparison with State-of-the-Art Methods
		3.4 Ablation Studies
		3.5 Visual Analysis
	4 Summary
	References
Sketch Image Style Transfer Based on Sketch Density Controlling
	1 Introduction
	2 Related Works
	3 Method
		3.1 Overall Pipeline
		3.2 Sketch Density Extraction and Encoding
		3.3 Style Encoder
		3.4 Density Controlled Style Transfer
	4 Data Construction
	5 Experiments
		5.1 Implementation Details
		5.2 Ablation Study
		5.3 Comparison with SOTA Methods
	6 Conclusion
	References
VAE-AD: Unsupervised Variational Autoencoder for Anomaly Detection in Hyperspectral Images
	1 Introduction
	2 Methodology
		2.1 Dimensionality Reduction Using PCA
		2.2 Background Reconstruction Using VAE
		2.3 Anomaly Map Construction
	3 Experimental Setups
		3.1 Dataset
		3.2 Evaluation Criteria
	4 Results and Discussion
		4.1 Analysis of Dimensionality Reduction Process
		4.2 Analysis of Detection Map Construction Using Gaussian Mixture Model
		4.3 Comparison with State of the Art Methods
	5 Conclusion
	References
DSE-Net: Deep Semantic Enhanced Network for Mobile Tongue Image Segmentation
	1 Introduction
	2 Relate Work
	3 Proposed Method
		3.1 Lightweight Feature Extraction Module
		3.2 Deep Semantic Enhanced Module
		3.3 Decoder
		3.4 Loss Function
	4 Experimental Results
		4.1 Dataset
		4.2 Evaluation Metrics
		4.3 Implementation Details
		4.4 Comparison with the State-of-the-Art
	5 Conclusion
	References
Efficient-Nets and Their Fuzzy Ensemble: An Approach for Skin Cancer Classification
	1 Introduction
	2 Related Works
	3 Dataset
	4 Methods
		4.1 Reward Function
		4.2 Ensemble: Choquet Fuzzy Integral
	5 Results
	6 Conclusion and Future Work
	References
A Framework for Software Defect Prediction Using Optimal Hyper-Parameters of Deep Neural Network
	1 Introduction
	2 Related Research
	3 Proposed Approach
		3.1 Dataset Imbalance Handling and Scaling
		3.2 Grid Search Based Hyper-Parameter Optimization Technique (GSBO)
	4 Experimental Setup
		4.1 Research Questions
	5 Results and Discussion
		5.1 Answers of RQs
	6 Conclusion and Future Scope
	References
Improved Feature Fusion by Branched 1-D CNN for Speech Emotion Recognition
	1 Introduction
	2 Related Work
	3 Motivation
	4 Methodology
		4.1 Proposed Improved Method
		4.2 Data Augmentation
		4.3 Feature Extraction
		4.4 1-D CNN Architecture
		4.5 Model Evaluation
	5 Datasets
	6 Results and Discussions
		6.1 Environment and Implementation Details
		6.2 Loss and Accuracy Curves for Emo-DB and TESS
		6.3 Classification Report for Emo-DB and TESS
		6.4 Comparative Analysis for Emo-DB and TESS
	7 Conclusion
	References
A Multi-modal Graph Convolutional Network for Predicting Human Breast Cancer Prognosis
	1 Introduction
	2 Proposed Work
		2.1 Data Set
		2.2 Graph Convolutional Network(GCN)
		2.3 Steps of Proposed Technique (MGCN)
		2.4 Experimental Setup
		2.5 Objective Function
	3 Experimental Results
		3.1 Evaluation Measures
		3.2 Results Using 10 Fold Cross-Validation Setting
		3.3 Comparison with Other Prediction Methods
		3.4 Discussion
	4 Conclusion
	References
Anomaly Detection in Surveillance Videos Using Transformer Based Attention Model
	1 Introduction
	2 Related Work
	3 Proposed Method
		3.1 Stage 1 (Feature Extraction)
		3.2 Stage 2 (Attention Layer)
		3.3 Stage 3 (Anomaly Detection)
	4 Experiments
		4.1 Dataset Description
		4.2 Evaluation Metric
		4.3 Implementation Details
		4.4 Result Analysis
	5 Conclusion
	References
Change Detection in Hyperspectral Images Using Deep Feature Extraction and Active Learning
	1 Introduction
	2 Related Research
	3 Proposed Method
		3.1 Unsupervised Pre-training of Sliding Convolutional Autoencoder
		3.2 Active Re-training of Twin Encoder Change Detector (TECD) Model
	4 Results
		4.1 Dataset Description
		4.2 Experiments
	5 Conclusion
	References
TeethU2Net: A Deep Learning-Based Approach for Tooth Saliency Detection in Dental Panoramic Radiographs
	1 Introduction
	2 Methods
		2.1 Network Architecture
		2.2 Loss Function
		2.3 Implementation Details
		2.4 Dataset
		2.5 Evaluation Metrics
	3 Results
		3.1 Quantitative Analysis
		3.2 Qualitative Analysis
	4 Conclusion and Future Work
	References
The EsnTorch Library: Efficient Implementation of Transformer-Based Echo State Networks
	1 Introduction
	2 Related Works
	3 ESNs for Text Classification
	4 EsnTorch
		4.1 Dataset
		4.2 Model
		4.3 Training
		4.4 Evaluation
		4.5 Deep ESNs
	5 Conclusion
	References
Wine Characterisation with Spectral Information and Predictive Artificial Intelligence
	1 Introduction
	2 Related Work
	3 The Color of GJ/WINE
	4 Grape Juice Tasting
		4.1 Data Collection
		4.2 Prediction Results
	5 Juice (or Grape) Origin Prediction
		5.1 Data Collection
		5.2 Prediction Results
	6 Wavelength Importance
	7 Conclusions
	References
MRCE: A Multi-Representation Collaborative Enhancement Model for Aspect-Opinion Pair Extraction
	1 Introduction
	2 Related Works
	3 Our Approach
		3.1 Task Definition
		3.2 Network Architecture
		3.3 Recurrent Interactive Units
		3.4 Joint Training
		3.5 Inference Layer
	4 Experiments
		4.1 Datasets and Experimental Settings
		4.2 Baselines
		4.3 Main Results
		4.4 Model Analysis
		4.5 Case Study
	5 Conclusion
	References
Diverse and High-Quality Data Augmentation Using GPT for Named Entity Recognition
	1 Introduction
	2 Related Work
		2.1 Data Augmentation
		2.2 Prompt Learning
	3 Our Method
		3.1 Data Preprocessing
		3.2 Data Generation
	4 Experiments
		4.1 Dataset
		4.2 Evaluation
		4.3 Main Result
	5 Analysis
	6 Conclusion and Future Work
	References
Transformer-Based Original Content Recovery from Obfuscated PowerShell Scripts
	1 Introduction
	2 Related Works
	3 Deobfuscation
	4 Dataset
	5 Model
	6 Training
	7 Results
	8 Conclusion
	References
A Generic Enhancer for Backdoor Attacks on Deep Neural Networks
	1 Introduction
	2 Preliminaries
		2.1 Backdoor Attacks
		2.2 Backdoor Defenses
	3 Methodology
		3.1 Threat Model
		3.2 Intuition
		3.3 Proposed Backdoor Enhancer
	4 Experiments
		4.1 Experiment Setting
		4.2 Results and Analysis
	5 Ablation Studies
		5.1 Effect on STRIP FAR by Varying FRR
		5.2 Effect of Varying Label Smoothing Factor
		5.3 Effect of Varying Activation Suppression Factor
	6 Conclusion
	References
Attention Based Twin Convolutional Neural Network with Inception Blocks for Plant Disease Detection Using Wavelet Transform
	1 Introduction
	2 Related Work
	3 Proposed Work
	4 Results and Discussion
		4.1 Performance Evaluation
		4.2 Ablation Study
	5 Conclusions and Future Work
	References
A Medical Image Steganography Scheme with High Embedding Capacity to Solve Falling-Off Boundary Problem Using Pixel Value Difference Method
	1 Introduction
	2 PVD Method
	3 Proposed Methodology
	4 Embedding Algorithm
	5 Extraction Algorithm
		5.1 Experimental Results
	6 Comparison
	7 Conclusion
	References
Deep Ensemble Architecture: A Region Mapping for Chest Abnormalities
	1 Introduction
	2 Related Work
	3 Data Exploration
		3.1 Weakly Supervised Dataset
		3.2 Database of Region Mapping
	4 Deep Ensemble Architecture
		4.1 Localization
		4.2 Region Mapping
	5 Experiments and Results
	6 Conclusion
	References
Privacy-Preserving Federated Learning for Pneumonia Diagnosis
	1 Introduction
	2 Literature Survey
	3 Methodology
		3.1 Federated Learning
		3.2 Model Architecture
		3.3 Dataset Preprocessing
	4 Results and Discussion
		4.1 Data Analysis
		4.2 Performance Analysis
	5 Conclusion and Future Scope
	References
Towards Automated Segmentation of Human Abdominal Aorta and Its Branches Using a Hybrid Feature Extraction Module with LSTM
	1 Introduction
	2 Method
		2.1 Patch Embedding and Downsampling Module
		2.2 Hybrid Feature Module
		2.3 Decoder Part
		2.4 Loss
		2.5 Performance Evaluation Metrics
	3 Results and Discussion
		3.1 Clinical Data and Hardware
		3.2 Experiment Results and Discussion
	4 Conclusion
	References
P-LSTM: A Novel LSTM Architecture for Glucose Level Prediction Problem
	1 Introduction
	2 Related Work
	3 Methodology
		3.1 A Novel Activation Function: Motivation
		3.2 Structure of P-LSTM
		3.3 Handling the Parameter of the Elliott Activation
		3.4 Causality Analysis
	4 Dataset and Experiment
		4.1 Dataset
		4.2 LSTM Model and Evaluation
		4.3 Results
	5 Conclusion
	References
Wide Ensemble of Interpretable TSK Fuzzy Classifiers with Application to Smartphone Sensor-Based Human Activity Recognition
	1 Introduction
	2 On Interpretable Zero-Order TSK Fuzzy Classifiers
	3 The Proposed Classifier WEIFC
		3.1 The Proposed Wide Structure
		3.2 B. Learning Algorithm and Computational Complexity
	4 Experimental Results
		4.1 Datasets and Parameter Settings
		4.2 Results and Discussions
	5 Conclusion
	References
Prediction of the Facial Growth Direction: Regression Perspective
	1 Introduction
	2 Related Literature
	3 Facial Growth Prediction
		3.1 Face Normalization and Problem Definition
		3.2 Summary of Our Previous Results
	4 Regression Analysis
		4.1 Experimental Setup
		4.2 How Well Cephalogram-Derived Features Predict FG Changes?
		4.3 Further Analysis
	5 Conclusions
	References
A Methodology for the Prediction of Drug Target Interaction Using CDK Descriptors
	1 Introduction
	2 Computation Approaches to DTI Predictions
	3 Materials and Methods
		3.1 Dataset Description
		3.2 Drug Feature Representation
		3.3 Protein Feature Representation
	4 Proposed Methodology
		4.1 CatBoost
	5 Evaluation Parameters
	6 Performance Evaluation
		6.1 Comparison with Previous Methods
	7 Conclusion
	References
PSSM2Vec: A Compact Alignment-Free Embedding Approach for Coronavirus Spike Sequence Classification
	1 Introduction
	2 Related Work
	3 Proposed Approach
		3.1 PSSMFreq2Vec
		3.2 PSSM2Vec
	4 Experimental Setup
		4.1 Evaluation Metrics
		4.2 Baseline Models
	5 Results and Discussion
		5.1 Classification Results
		5.2 Evaluating t-SNE
		5.3 Statistical Analysis
	6 Conclusion
	References
An Optimized Hybrid Solution for IoT Based Lifestyle Disease Classification Using Stress Data
	1 Introduction
	2 Related Work
	3 Material and Methods
		3.1 Dataset Description and Hardware Used for Data Acquisition
		3.2 Data Preparation
		3.3 Implementation of the Proposed Model
	4 Result and Discussion
		4.1 Accuracy of Various Classifier Using Feature Selection Techniques
	5 Conclusion and Future Scope
	References
A Deep Concatenated Convolutional Neural Network-Based Method to Classify Autism
	1 Introduction
	2 Related Work
	3 Proposed Model
		3.1 Dataset
		3.2 Methodology
	4 Result and Discussion
	5 Conclusions
	References
Deep Learning-Based Human Action Recognition Framework to Assess Children on the Risk of Autism or Developmental Delays
	1 Introduction
	2 Material and Methods
		2.1 Objective
		2.2 Dataset
		2.3 Methods
	3 Experiments and Results
		3.1 Development of DL Models
		3.2 Validation of DL Models
	4 Discussion
	5 Conclusion and Future Scope
	References
Dynamic Convolutional Network for Generalizable Face Anti-spoofing
	1 Introduction
	2 Related Work
		2.1 Face Anti-spoofing
		2.2 Dynamic Convolutional Layer
	3 Methodology
		3.1 Overview
		3.2 Dynamic Domain Convolution Generator
		3.3 Asymmetric Center Mining
	4 Experiments
		4.1 Experimental Settings
		4.2 Comparison Results
		4.3 Ablation Study
		4.4 Visualization and Analysis
	5 Conclusion
	References
Challenges of Facial Micro-Expression Detection and Recognition: A Survey
	1 Introduction
	2 Major Challenges During Micro-Expression Recognition
		2.1 Climate Deviation
		2.2 The Unconstrained and Refined Motion of Facial Movement
		2.3 Imbalanced Dataset for Normal Situations
		2.4 In the Implementation of Deep Learning (High Level Representation)
	3 Micro-Expression Recognition Process
		3.1 Preprocessing
		3.2 Face Detection
		3.3 Face Registration
		3.4 Motion Magnification
		3.5 Temporal Normalization
	4 Feature Extraction
		4.1 Feature Detection
		4.2 Feature Representation
	5 Limitations of Existing Methods of Feature Descriptors
	6 Conclusion
	References
Biometric Iris Identifier Recognition with Privacy Preserving Phenomenon: A Federated Learning Approach
	1 Introduction
	2 Literature Review
	3 Methodology
		3.1 Dataset
		3.2 Data Preparation
		3.3 Privacy Preserving Model Training Approach: Federated Learning
		3.4 Feature Extraction
		3.5 Classification
	4 Results and Discussion
	5 Conclusion
	References
Traffic Flow Forecasting Using Attention Enabled Bi-LSTM and GRU Hybrid Model
	1 Introduction
	2 Related Work
	3 Proposed Approach
	4 Performance Evaluation
	5 Conclusion and Future Work
	References
Commissioning Random Matrix Theory and Synthetic Minority Oversampling Technique for Power System Faults Detection and Classification
	1 Introduction
	2 Literature Survey
	3 Approach for Contingency Data Generation and Detection
		3.1 Application of RMT and SMOTE
	4 Proposed Methodology: Detection and Classification of Line Faults
	5 Experiment and Analysis
		5.1 Classification of Power System Faults Using Binary Classifier
		5.2 Classification of Power System Faults Using Rule Based Decision Tree
	6 Conclusion
	References
Deep Reinforcement Learning with Comprehensive Reward for Stock Trading
	1 Introduction
	2 Method
		2.1 Policy Network Model
		2.2 Policy Network Optimization by RL
		2.3 Proposal Reward Function
		2.4 Sentiment Analysis of Stock
	3 Experiment
		3.1 Evaluation Criterion
		3.2 Implentation Details
		3.3 Results and Discussion
	4 Conclusion
	References
Deep Learning Based Automobile Identification Application
	1 Introduction
	2 Convolutional Neural Network and ITS Extensions
		2.1 Convolutional Neural Network
		2.2 Deep Residual Network
	3 Car Classification and Detection
	4 Models
	5 Experiment and Discussion
		5.1 Experimental Setup
		5.2 Experimental Results
		5.3 Evaluation and Application
	6 Conclusion
	References
Automatic Firearm Detection in Images and Videos Using YOLO-Based Model
	1 Introduction
		1.1 Yolo Algorithm
	2 Related Work
	3 Methodology
		3.1 Analysis
	4 Conclusion
	References
Author Index