Neural Information Processing: 30th International Conference, ICONIP 2023, Changsha, China, November 20–23, 2023, Proceedings, Part X (Communications in Computer and Information Science)

✍ Scribed by Biao Luo (editor), Long Cheng (editor), Zheng-Guang Wu (editor), Hongyi Li (editor), Chaojie Li (editor)

• Publisher: Springer
• Year: 2023
• Tongue: English
• Leaves: 629
• Category: Library

✦ Synopsis

The nine-volume set constitutes the refereed proceedings of the 30th International Conference on Neural Information Processing, ICONIP 2023, held in Changsha, China, in November 2023.
The 1274 papers presented in the proceedings set were carefully reviewed and selected from 652 submissions.
The ICONIP conference aims to provide a leading international forum for researchers, scientists, and industry professionals who are working in neuroscience, neural networks, deep learning, and related fields to share their new ideas, progress, and achievements.

✦ Table of Contents

Preface
Organization
Contents – Part X
Human Centred Computing
Paper Recommendation with Multi-view Knowledge-Aware Attentive Network
1 Introduction
2 Problem Formulation
3 The Proposed Framework
3.1 Knowledge Graph Construction
3.2 Multi-view Knowledge-Aware Attentive Layers
3.3 Information Aggregation and High-Order Propagation
3.4 Prediction and Optimization
4 Experiments
4.1 Experiment Settings
4.2 Experiment Results and Analysis
4.3 Comparison of Ablation Experiments
5 Conclusion and Future Work
References
Biological Tissue Sections Instance Segmentation Based on Active Learning
1 Introduction
2 Methodology
2.1 Loss Prediction Module
2.2 Uncertainty of Instance Segmentation Result
2.3 Active Learning Strategy for Instance Segmentation
3 Experimental Results
3.1 Dataset
3.2 Results
4 Conclusion
References
Research on Automatic Segmentation Algorithm of Brain Tumor Image Based on Multi-sequence Self-supervised Fusion in Complex Scenes
1 Introduction
2 Methods
2.1 Motivation for Model Design
2.2 Multi-sequence Self-supervised Fusion Segmentation Network
2.3 Backbone: Bi-ConvLSTM
2.4 Neck: Transformer+MLP
2.5 Head: Self-supervised Multi-sequence Integration
3 Experiment and Results
3.1 Dataset
3.2 Experimental Setup
3.3 Results
3.4 Ablation Experiment
4 Comparison with Existing Methods
4.1 Traditional Image Processing Methods
4.2 Methods Based on Deep Learning
4.3 Our Proposed Method
5 Conclusions
6 Outlook and Prospects Outlook
References
An Effective Morphological Analysis Framework of Intracranial Artery in 3D Digital Subtraction Angiography
1 Introduction
2 Related Work
3 Method
3.1 RGFNet for Segmentation of Intracranial Artery in 3D-DSA
3.2 Automatic Morphological Analysis Algorithm of Intracranial Artery
4 Experiments
4.1 Datasets
4.2 Implementation Details
4.3 Segmentation Results
4.4 Ablation Study of RGFNet
4.5 Morphological Analysis Results
5 Conclusion
References
Effective Domain Adaptation for Robust Dysarthric Speech Recognition
1 Introduction
2 Proposed Domain Adaptation Methods
2.1 Data Preprocessing
2.2 Domain-Adapted Transformer
3 Experiments
3.1 Data Description
3.2 Experimental Setup
3.3 Evaluation on Speaker-Dependent DSR
3.4 Evaluation on Speaker-Independent DSR
4 Conclusions
References
TCNet: Texture and Contour-Aware Model for Bone Marrow Smear Region of Interest Selection
1 Introduction
2 Related Work
2.1 Bone Marrow Smear Cell Morphology
2.2 Bone Marrow Smear ROI Selection
3 Method
3.1 Overall Model Structure
3.2 Texture Deep Supervision Module (TDSM)
3.3 Contour Deep Supervision Module (CDSM)
3.4 Loss Function
4 Experiment
4.1 Experimental Setup
4.2 Comparison Experiments
4.3 Ablation Experiments
5 Conclusion
References
Diff-Writer: A Diffusion Model-Based Stylized Online Handwritten Chinese Character Generator
1 Introduction
2 Related Work
2.1 Autoregressive Methods for Online Handwriting Generation
2.2 Online Chinese Character Generation
2.3 Non-autoregressive Diffusion Model
3 Methods
3.1 Denoising Diffusion Probabilistic Models
3.2 Data Representation
3.3 Diff-Writer Model
3.4 Training Process
3.5 Generation Process
4 Experiment and Results
4.1 Dataset
4.2 Implementation Details
4.3 Evaluations
5 Conclusion and Future Work
References
EEG Epileptic Seizure Classification Using Hybrid Time-Frequency Attention Deep Network
1 Introduction
2 The Construction of TFACBNet
2.1 Time-Frequency Attention Module
2.2 Hybrid Deep Network Architecture
3 Experiment and Results
3.1 Database
3.2 Evaluation Metric
3.3 Experimental Results
4 Conclusion
References
A Feature Pyramid Fusion Network Based on Dynamic Perception Transformer for Retinal OCT Biomarker Image Segmentation
1 Introduction
2 Proposed Method
2.1 Feature Pyramid Fusion Module
2.2 Dynamic Scale Transformer Module
2.3 Loss Function
3 Experiment
3.1 Dataset
3.2 Comparison Methods and Evaluation Metrics
3.3 Results
3.4 Ablation Experiment
4 Conclusion
References
LDW-RS Loss: Label Density-Weighted Loss with Ranking Similarity Regularization for Imbalanced Deep Fetal Brain Age Regression
1 Introduction
2 Method
2.1 Label Density-Weighted Loss
2.2 Ranking Similarity Regularization
3 Experiments and Results
3.1 Experimental Settings
3.2 Comparison with Other State-of-the-Art Methods
3.3 Correlation and Agreement Analysis
3.4 Ablation Study
4 Conclusion
References
Segment Anything Model for Semi-supervised Medical Image Segmentation via Selecting Reliable Pseudo-labels
1 Introduction
2 Method
2.1 The Selection of Reliable Pseudo Label (Assisted by SAM)
2.2 Re-training the Model with Reliable Pseudo-labels (Taking SDA-MT as an Example)
3 Experiment and Results
3.1 Dataset
3.2 Implementation Details and Evaluation Metrics
3.3 Results
4 Conclusion
References
Aided Diagnosis of Autism Spectrum Disorder Based on a Mixed Neural Network Model
1 Introduction
2 Preliminaries
2.1 Datasets and Preprocessing
2.2 The Calculation of Functional Connectivity Matrix
2.3 Global Features Selection
2.4 Functional Gradient
3 Methodology and Materials
3.1 The Architecture of the Proposed Model
3.2 The Comparison Models
3.3 Experimental Setting and Evaluation Metrics
4 Experiment Results and Discussion
4.1 Brain Structure Analysis
5 Conclusion
References
A Supervised Spatio-Temporal Contrastive Learning Framework with Optimal Skeleton Subgraph Topology for Human Action Recognition
1 Introduction
2 Relation Work
2.1 Skeleton-Based Action Recognition
2.2 Contrastive Learning
3 Method
3.1 Supervised Spatio-Temporal Contrastive Learning Framework
3.2 Supervised Contrastive Loss
3.3 Optimal Skeleton Subgraph Topology Determination
4 Experiments
4.1 Datasets
4.2 Experimental Configuration
4.3 Ablation Study
4.4 Comparison with Other Methods
5 Conclusion
References
Multi-scale Feature Fusion Neural Network for Accurate Prediction of Drug-Target Interactions
1 Introduction
2 Materials and Methods
2.1 Drug Representation Learning Branch
2.2 Protein Representation Learning Branch
2.3 Feature Fusion and Prediction Module
3 Experiments and Results
3.1 Datasets
3.2 Implementation Details
3.3 DTI Prediction
3.4 Binding Affinity Prediction
4 Conclusion
References
GoatPose: A Lightweight and Efficient Network with Attention Mechanism
1 Introduction
2 Related Work
3 Approach
3.1 Parallel Multi-branch Architecture
3.2 Model Lightweighting
3.3 Adaptive Attention Mechanism
3.4 GoatPose
4 Experiments
4.1 Dataset
4.2 Setting
4.3 Results
4.4 Deployment
5 Conclusion
References
Sign Language Recognition for Low Resource Languages Using Few Shot Learning
1 Introduction
2 Related Work
3 SSL50 Dataset
4 Proposed ProtoSign Framework
4.1 Skeleton Locations Extraction
4.2 Sign Video Encoder
5 Experimental Study
5.1 Datasets
5.2 Implementation Details
5.3 Experimental Results
6 Conclusion
References
T Cell Receptor Protein Sequences and Sparse Coding: A Novel Approach to Cancer Classification
1 Introduction
2 Related Work
3 Proposed Approach
3.1 Incorporating Domain Knowledge
3.2 Cancer Types and Immunogenetics
3.3 Algorithmic Pseudocode and Flowchart
4 Experimental Setup
4.1 Dataset Statistics
4.2 Data Visualization
5 Results and Discussion
6 Conclusion
References
Weakly Supervised Temporal Action Localization Through Segment Contrastive Learning
1 Introduction
2 Related Work
3 Method
3.1 Feature Extraction and Embedding
3.2 Action Proposal
3.3 Inter- and Intra-segment Contrastive Learning
3.4 Training Objectives
3.5 Inference
4 Experiments
4.1 Experimental Setups
4.2 Comparison with the State-of-the-Arts
4.3 Ablation Study
4.4 Qualitative Results
5 Conclusion
References
S-CGRU: An Efficient Model for Pedestrian Trajectory Prediction
1 Introduction
2 Related Work
2.1 Complex-Valued Neural Network
2.2 Recurrent Neural Networks
2.3 Human-Human Interactions
3 Method
3.1 Problem Definition
3.2 Architecture Overview
3.3 Complex Gated Recurrent Unit
3.4 Exploring Pedestrians' Field of Vision in Different Environments
3.5 Loss Functions
4 Experiments
4.1 Dataset and Evaluation Metrics
4.2 Baselines
4.3 Evaluations on ADE/FDE Metrics
4.4 Model Parameter Amount and Inference Speed
4.5 The Range of Visual Focus of Pedestrians While Walking in Different Scenarios
5 Conclusions and Future Works
References
Prior-Enhanced Network for Image-Based PM2.5 Estimation from Imbalanced Data Distribution
1 Introduction
2 Related Work
3 Methodology
3.1 Estimation of DC-IS Maps
3.2 PE Network Architecture
3.3 Histogram Smoothing Algorithm
4 Experiments
4.1 Experiment Setup
4.2 Comparison with the SOTA Methods
4.3 Ablation Studies
4.4 Examples on Various Shooting Scenes
5 Conclusion
References
Dynamic Data Augmentation via Monte-Carlo Tree Search for Prostate MRI Segmentation
1 Introduction
2 Methodology
2.1 Search Space
2.2 Tree Construction
2.3 Tree Updating and Pruning
2.4 Tree Sampling
3 Implementation and Experiments
3.1 Datasets and Implementation Details
3.2 Experimental Results
4 Conclusion
References
Language Guided Graph Transformer for Skeleton Action Recognition
1 Introduction
2 Related Works
3 Method
3.1 Skeleton Data Preprocessing
3.2 GCN
3.3 Graph Transformer
3.4 Text Feature Module
4 Experiments
4.1 Dataset
4.2 Implementation Details
4.3 Ablation Study
4.4 Comparison with State-of-the-Art Methods
5 Conclusion
References
A Federated Multi-stage Light-Weight Vision Transformer for Respiratory Disease Detection
1 Introduction
2 Proposed Framework
2.1 Federated Learning
2.2 Light-Weight Vision Transformer
3 Experimental Dataset and Results
3.1 Classification Dataset
3.2 Results
3.3 Results on Centralized Data
3.4 Results on Federated Learning (IID and Non-IID Data Distribution)
3.5 Communication Time and Client Memory Requirement
3.6 Infection Localization and Error Analysis
4 Conclusion and Future Work
References
Curiosity Enhanced Bayesian Personalized Ranking for Recommender Systems
1 Introduction
2 Related Work
3 Notations and Problem Definition
4 Curiosity Enhanced Bayesian Personalized Ranking
4.1 Model Assumption
4.2 Modeling of Curiosity
4.3 Learning the CBPR
4.4 Item Recommendation
5 Experiments and Results Analysis
5.1 Experimental Settings
5.2 Recommendation Performance
5.3 Parameters Analysis
6 Conclusion
References
Modeling Online Adaptive Navigation in Virtual Environments Based on PID Control
1 Introduction
1.1 Cybersickness Evaluation
1.2 Adaptive Navigation
2 Adaptive Navigation Design
2.1 PID Controller
2.2 1D Convolutional Neural Network
2.3 Formulation of Adaptive Navigation
3 Data Collection and Parameters Computation
3.1 Data Acquisition
3.2 Model Architecture
3.3 Computing the Adaptive Coefficients
3.4 Results
4 Discussion
5 Conclusion
References
Lip Reading Using Temporal Adaptive Module
1 Introduction
2 The Proposed Work
2.1 Lip Reading Models Based on TAM-Net Models
2.2 The Overview of TAM-Net
2.3 Partially Dense TCN
3 Experiments
3.1 Datasets and Implementation Details
3.2 Evaluation of TAM
4 Conclusion
References
AudioFormer: Channel Audio Encoder Based on Multi-granularity Features
1 Introduction
2 Related Work
2.1 Feature Engineering
2.2 High-Order Feature Representation
3 Methodology
3.1 Problem Definition
3.2 Multi-granularity Features Extraction
3.3 Channel Audio Encoder Model Structure
3.4 Feature Encoder
4 Experiments and Dataset
4.1 Dataset Introduction
4.2 Evaluation Standard
4.3 Compared Baselines
4.4 Experimental Results
4.5 Real Scene Verification
5 Ablation Experiments
5.1 Channel Audio Encoder Ablation Experiment
5.2 Multi-granularity Features Ablation Experiment
6 Discussion
7 Conclusion
References
A Context Aware Lung Cancer Survival Prediction Network by Using Whole Slide Images
1 Introduction
2 Related Work
3 The Proposed CA-SurvNet
3.1 Pre-training the WSI_Encoder
3.2 The Network Architecture of CA-SurvNet
4 Experiment and Analysis
4.1 Dataset and Experiment Setup
4.2 Comparison with State-of-the-Arts
4.3 Ablation Study
5 Conclusion
References
A Novel Approach for Improved Pedestrian Walking Speed Prediction: Exploiting Proximity Correlation
1 Introduction
2 Related Work
2.1 Pedestrian Walking Speed Analysis and Forecast
2.2 Time Series Forecasting
3 Methodology
3.1 Problem Definition
3.2 Framework Overview
3.3 Time Series Forecast Module
3.4 Weighted Average Forecast Module
3.5 Fusion Prediction Module
4 Experiment
4.1 Dataset
4.2 Correlation Verification
4.3 Hyperparameter Settings
5 Results
5.1 Correlation Verification
5.2 Model Performance
6 Conclusion
References
MView-DTI: A Multi-view Feature Fusion-Based Approach for Drug-Target Protein Interaction Prediction
1 Introduction
2 Methodology
2.1 Drug Feature Extraction
2.2 Protein Feature Extraction
2.3 Mutual Feature Learning
3 Experiment
3.1 Dataset
3.2 Results
3.3 Robustness Experiment
3.4 Ablation Study
4 Conclusions
References
User Multi-preferences Fusion for Conversational Recommender Systems
1 Introduction
2 Related Work
2.1 Conversational Recommender Systems
2.2 Wide&Deep
3 Preliminary
4 Method
4.1 User's Own Preference Learner
4.2 Third-Party Information Learner
4.3 Attentive Wide&Deep Modeling
4.4 Optimization
5 Experiment
5.1 Experimental Setting
5.2 Evaluation on Recommendation Task
5.3 Evaluation on Conversation Task
5.4 Ablation Study
6 Conclusion and Future Work
References
Debiasing Medication Recommendation with Counterfactual Analysis
1 Introduction
2 Problem Formulation
3 The CAMeR
3.1 Historical Score Computation
3.2 Medication Generate Network
3.3 Counterfactual Debiasing Framework
3.4 Training Objective
4 Experiments
4.1 Dataset
4.2 Baselines and Metrics
4.3 Results
4.4 Ablation Study
4.5 Case Study
5 Related Work
6 Conclusion
References
Early Detection of Depression and Alcoholism Disorders by EEG Signal
1 Introduction
2 Materials and Method
2.1 Used Databases
2.2 Empirical Wavelet Transform
2.3 Feature Extraction
2.4 Feature Selection
2.5 Classifiers
3 Results and Discussion
4 Conclusion
References
Unleash the Capabilities of the Vision-Language Pre-training Model in Gaze Object Prediction
1 Introduction
2 Related Work
2.1 Gaze Following
2.2 Adapter for LLM
3 Method
3.1 EdgeCLIP
3.2 Gaze Prediction
3.3 Object Detection
3.4 Regulatory Loss
4 Experiments
4.1 Setups
4.2 Comparison with State-of-the-Arts
4.3 Ablation Studies
4.4 Qualitative Analysis
5 Conclusion
References
A Two-Stage Network for Segmentation of Vertebrae and Intervertebral Discs: Integration of Efficient Local-Global Fusion Using 3D Transformer and 2D CNN
1 Introduction
2 Methods
2.1 Overall Architecture Design
2.2 3D Coarse Segmentation Stage
2.3 2D Refinement Segmentation Stage
2.4 Graph Convolution Module
3 Experiments
3.1 Dataset
3.2 Implementation Details
3.3 Evaluation Metrics
3.4 Experiment Results
3.5 Ablation Study
3.6 Effect of the Two-Stage Framework
3.7 Effect of the GCM
4 Conclusion
References
Integrating Multi-view Feature Extraction and Fuzzy Rank-Based Ensemble for Accurate HIV-1 Protease Cleavage Site Prediction
1 Introduction
2 Proposed Work
2.1 Data Set
2.2 Feature Extraction
2.3 Fuzzy Rank-Based Ensemble
2.4 Steps of Proposed Technique
3 Experimental Results
4 Discussion
5 Conclusion
References
KSHFS: Research on Drug-Drug Interaction Prediction Based on Knowledge Subgraph and High-Order Feature-Aware Structure
1 Introduction
2 Method
2.1 Extracting Knowledge Graph Subgraph
2.2 Feature Information Perception Module
2.3 Binary DDI Prediction
2.4 Multi-relational DDIs Prediction
3 Experimental Setup and Result Analysis
3.1 Experimental Setup
3.2 Results Analysis
3.3 Ablation Study
3.4 Parameter Discussion
4 Conclusion
References
Self-supervised-Enhanced Dual Hierarchical Graph Convolution Network for Social Recommendation
1 Introduction
2 Related Work
3 Preliminaries
4 Methodology
4.1 Link Encoded-Weight Balanced Message Passing Paradigm
4.2 Dual Hierarchical Self-supervised Enhancement Based on Hypergraph
4.3 Model Training
5 Experiments
5.1 Experimental Settings
5.2 Performance Comparison
5.3 Ablation Study of SSHGCN Framework
6 Conclusion
References
Dynamical Graph Echo State Networks with Snapshot Merging for Spreading Process Classification
1 Introduction
2 Preliminaries
3 The Proposed Model
3.1 The Merged Snapshot Converter
3.2 The Multiple-Reservoir Encoder
3.3 The Linear Classifier
4 The Analysis of the Computational Complexity
5 Experiments
5.1 Descriptions of Datasets
5.2 Tested Models and Experimental Settings
5.3 Evaluation Metrics
5.4 Experimental Results
6 Discussion
References
Trajectory Prediction with Contrastive Pre-training and Social Rank Fine-Tuning
1 Introduction
2 Related Work
3 Our Approach
3.1 Contrastive History-Prediction Learning
3.2 Differentiable Social Interaction Ranking
3.3 Bi-Gaussian Regression Loss
3.4 Two-Stage Multi-task Training Objective
4 Evaluation Metrics
4.1 Standard ADE and FDE
4.2 Surrogate Social Distance Accuracy (SDA)
5 Experiments
5.1 Our Approaches and Baselines
5.2 ADE and FDE Results
5.3 SDA Results
5.4 Ablation Study of Model Adaptivity
5.5 Visualizations of Ranking
6 Conclusion
References
Three-Dimensional Medical Image Fusion with Deformable Cross-Attention
1 Introduction
2 Methods
2.1 Overview of Proposed Method
2.2 Deformable Cross Feature Blend (DCFB)
3 Experiments
3.1 Data Preparation and Evaluation Metrics
3.2 Implementation Details
3.3 Results and Analysis
4 Conclusion
References
Handling Class Imbalance in Forecasting Parkinson's Disease Wearing-off with Fitness Tracker Dataset
1 Introduction
2 Methodology
2.1 Data Collection
2.2 Data Processing
2.3 Class Imbalance Techniques, Experiments
3 Results
3.1 How Can Data Collection Be Improved for Nursing Care Staff to Increase the Number of Wearing-off Labels in the Dataset?
3.2 Which Class Imbalance Techniques Performed Well in Forecasting Wearing-Off in the Next Hour?
3.3 How Does Reweighting the Forecast Probabilities Due to Resampling Affect the Forecasting Models?
4 Discussions
5 Conclusion
References
Real-Time Instance Segmentation and Tip Detection for Neuroendoscopic Surgical Instruments
1 Introduction
2 Related Work
3 Approach
3.1 Framework
3.2 Data Augmentation
4 Results and Discussion
4.1 Datasets and Metrics
4.2 Results
5 Conclusion
References
Spatial Gene Expression Prediction Using Hierarchical Sparse Attention
1 Introduction
2 Related Work
3 Method
3.1 Preliminaries
3.2 Hierarchical Sparse Attention Network (HSATNet)
4 Experiments
4.1 Datasets
4.2 Experimental Set-Up
4.3 Experimental Results
4.4 Ablation Study
5 Conclusion
References
Author Index