Advanced Intelligent Computing Technology and Applications : 20th International Conference, ICIC 2024, Tianjin, China, August 5–8, 2024, Proceedings, Part IV / edited by De-Shuang Huang, Chuanlei Zhang, Wei Chen.

Format:

Book

Contributor:

Huang, De-Shuang, editor.

Zhang, Chuanlei, editor.

Chen, Wei, editor.

Series:

Lecture Notes in Computer Science, 1611-3349 ; 14865

Language:

English

Subjects (All):

Computational intelligence.

Computer networks.

Machine learning.

Application software.

Computational Intelligence.

Computer Communication Networks.

Machine Learning.

Computer and Information Systems Applications.

Local Subjects:

Computational Intelligence.

Computer Communication Networks.

Machine Learning.

Computer and Information Systems Applications.

Physical Description:

1 online resource (516 pages)

Edition:

1st ed. 2024.

Place of Publication:

Singapore : Springer Nature Singapore : Imprint: Springer, 2024.

Summary:

This 13-volume set LNCS 14862-14874 constitutes - in conjunction with the 6-volume set LNAI 14875-14880 and the two-volume set LNBI 14881-14882 - the refereed proceedings of the 20th International Conference on Intelligent Computing, ICIC 2024, held in Tianjin, China, during August 5-8, 2024. The total of 863 regular papers were carefully reviewed and selected from 2189 submissions. This year, the conference concentrated mainly on the theories and methodologies as well as the emerging applications of intelligent computing. Its aim was to unify the picture of contemporary intelligent computing techniques as an integral concept that highlights the trends in advanced computational intelligence and bridges theoretical research with applications. Therefore, the theme for this conference was "Advanced Intelligent Computing Technology and Applications". Papers that focused on this theme were solicited, addressing theories, methodologies, and applications in science and technology.

Contents:

Intro

Preface

Organization

Contents - Part IV

Neural Networks

Multi-mode Graph Attention-Based Anomaly Detection on Attributed Networks

1 Introduction

2 Problem Definition

3 The MulGAT Approach

3.1 Network Encoder

3.2 Structure and Attribute Decoder

3.3 Anomaly Detection

3.4 Experimental Setup

3.5 Results Analysis

3.6 Parameter Analysis

3.7 Ablation Study

4 Conclusions

References

A Hierarchical Multi-scale Cortical Learning Algorithm for Time Series Forecasting

2 Related Work

2.1 CLA

2.2 Multi-scale Method

3 Hierarchical Multi-scale CLA

3.1 Hierarchical Multi-scale CLA Structure

3.2 Multi-scale Temporal Context Disentanglement

3.3 Adaptive Decoder for Prediction Aggregation

4 Experiment

4.1 Datasets

4.2 Comparison Algorithms

4.3 Evaluation Metric

4.4 Experimental Parameters

4.5 Time Series Forecasting

4.6 Epoch Evaluation

4.7 Comparison of Column Number Training

5 Conclusion

Knowledge Tracing with Contrastive Learning and Attention-Based Long Short-Term Memory Network

2 Model

2.1 Problem Formulation

2.2 Embedding

2.3 ALSTM

2.4 Prediction

2.5 CL Framework

3 Experiments and Analysis

3.1 Datasets

3.2 Baselines

3.3 Evaluation Metrics and Parameter Settings

3.4 Comparison of Performance (RQ1)

3.5 Comparison on Augmentation Methods (RQ2)

3.6 Influence of Model Components (RQ3)

3.7 Case Study

4 Conclusion

Multi-label Classification for Concrete Defects Based on EfficientNetV2

2.1 Attention Mechanism

3 Methodology

3.1 Spatial and Channel Reconstruction Attention Module

3.2 Spatial Pyramid Pooling-Fast

3.3 Asymmetric Loss

4 Experimental Study

4.1 Dataset

4.2 Evaluation Metrics.

4.3 Experiment Results and Analysis

3D Convolution Channel Compression for Stereo Matching

3 Method

3.1 Analysis of the Cost Aggregation Module

3.2 3D Convolution Channel Compression for Stereo Matching

3.3 Analysis of Computational Efficiency

4 Experiments and Results

4.1 Datasets and Evaluation Metrics

4.2 Experimental Setup

4.3 Channel Compression for PSMNet

4.4 Channel Compression for GwcNet

4.5 Channel Compression for CFNet

4.6 KITTI Benchmark

Designing Real-Time Neural Networks by Efficient Neural Architecture Search

2.1 Neural Network Design

2.2 Time Analysis of Neural Networks

3 Problem Formulation

4 RetNAS Framework

4.1 EVT-Based WCET Estimator

4.2 Constraint Schedule

4.3 Maximum Entropy-Based Search

5 Experiments

5.1 Experimental Settings

5.2 Evaluation for WCET Estimator

5.3 Comparison with Hand-Crafted Neural Architecture

5.4 Comparative Analysis with SOTA NAS Methods

6 Conclusion

Uncertainty-Driven Multi-scale Feature Fusion Network for Real-Time Image Deraining

2 Uncertainty-Driven Multi-scale Feature Fusion Network

2.1 Encoder-Decoder Deraining Network

2.2 Uncertainty Estimation

2.3 Loss Function

3 Experimental

3.1 Experimental Settings

3.2 Experimental Settings

3.3 Ablation Study

3.4 Computational Costs

RAY-Net: A Motorcycle Helmet Detection Method Integrated Auxiliary Correction

2 Related Works

3 Datasets

3.1 Open-Source Datasets

3.2 EMHDD

4 Methodology

4.1 RAY-Net

4.2 Detector

4.3 Recognizer

5.1 Settings

5.2 Comparison Experiments

5.3 Ablation Experiments.

6 Conclusions

Augmented Fuzzy Min-Max Neural Network Driven to Preprocessing Techniques and Space Search Optimization Algorithm

2 Preliminaries and Motivation

2.1 Fuzzy Min-Max Neural Network

2.2 Motivation of AFMNN

3 Architecture of AFMNN

3.1 Preprocessing Part by Using Information Gain (IG)

3.2 Hyperbox Generation Part

4 Design of AFMNN

5 Experimental Studies

5.1 Experiment I

5.2 Experiment II

5.3 Experiment III

5.4 Experiment IV

Building Change Detection Based on Fully Convolutional Network in High-Resolution Remote Sensing Images

3.1 Overall Structure

3.2 Feature Difference Enhancement (FDE) Module

3.3 Boundary Feature Compensation (BFC) Module

3.4 Multi-Scale Feature Aggregation (MSFA) Module

3.5 Change Guide Module (CGM)

3.6 Loss Function

4 Experimental and Analysis

4.3 Comparative Experiment

4.4 Ablation Experiments

Optimization of NUMA Aware DNN Computing System

2.1 Universal NUMA Optimization Technology

2.2 Framework Level NUMA Optimization Technology

3 DNN Computing Memory Access Latency Model and Optimization Ideas for NUMA Awareness

3.1 NUMA Memory Access Latency Model for DNN Computing

3.2 Optimization Problem Based on Latency Models

3.3 Optimization Ideas for NUMA Aware DNN Computing

4 Design of NUMA Aware DNN Computing System

4.1 Optimization of Memory Access Patterns Consistency

4.2 Batch-Level Parallel Task Allocation

4.3 Page-Aligned Memory Allocation and NUMA Aware Memory Pool

5 Experiments and Evaluation

5.1 Overall Evaluation

5.2 Evaluation of Each Layer

5.3 Evaluation of Scalability

6 Conclusion.

References

Pest-YOLO: A Lightweight Pest Detection Model Based on Multi-level Feature Fusion

2 Data Augmentation

3.1 Pest-YOLO Model Architecture

3.2 Ghost-CF Module

3.3 SERes Detection Heads

4 Experimental Analysis

4.1 Experimental Dataset

4.2 Experimental Environment

4.3 Comprehensive Experimental Results

Enhanced Tiny Object Detection in Aerial Images

2.1 Multi-scale Feature Fusion

2.2 Context-Sensitive Feature Fusion

3.1 Improved YOLOv5 Framework

3.2 Depth-Wise-Reshaping Module

3.3 Multi-branch Deep Stripe Attention Module

3.4 Decoupled-Head Module

4.1 Experimental Settings

4.2 Evaluation Metrics

4.3 Main Results

4.4 Qualitative Research

4.5 Ablation Studies

Strategic Reparameterization for Enhanced Inference in Imperfect Information Games: A Neural Network Approach

3 Methods

3.1 Enhanced Inferential Reasoning Framework via Reparameterization Method

3.2 Constructing a Normal Distribution Aligned with the Current Game Process

4 Experiments

4.1 Experimental Setup and Data Analysis Methods

4.2 Experimental Analysis of Military Chess

4.3 Experimental Analysis of Dark Chess

4.4 Experimental Analysis of Lord of Rings

TSMGAN-II: Generative Adversarial Network Based on Two-Stage Mask Transformer and Information Interaction for Speech Enhancement

1.1 Background

1.2 Challenges

1.3 Contributions

2 Proposed Method

2.1 Method Description

2.2 Generator

2.3 Attention Encoder

2.4 Two-Stage Mask Transformer

2.5 Dual-Feature Decoder with Information Interaction

2.6 Lightweight Model.

2.7 Discriminator

2.8 Loss Function

3 Experiment

3.1 Experimental Setup

3.2 Baseline

3.3 Ablation Experiment

3.4 General Experiment

Prompt-Based Event Temporal Relation Extraction with Contrastive Learning

2.1 Event Temporal Relation Extraction

2.2 Prompt-Based Learning

2.3 Contrastive Learning

3.1 Prompt Template and Verbalizer Design

3.2 Prompt-Based Supervised Contrastive Loss

3.3 Prompt-Based Tuning

4 Experiment Settings

4.2 Baselines

4.3 Experimental Results

4.4 Ablation Study

Subdomain Adaption Network Combining Cosine Distance and Angle Margin

2.1 Domain Adaptation

2.2 Large Margin Cosine Loss and Additive Angular Margin Loss

3.1 Maximum Mean Discrepancy

3.2 Combined Large Margin Cosine Loss and Additive Angular Margin Loss

3.3 Standardized Cosine Distance Loss

3.4 Network Structure and Cost Function

4.2 Setup

4.3 Results

4.5 Feature Visualization

4.6 Parameter Sensitivity Analysis

Self-attention-Based Dual-Branch Person Re-identification

3 Proposed Approach

3.1 Dual-Branch Network

3.2 Self-attention Model

4.1 Comparison with Existing Methods

4.2 Ablation Experiment

Open-Vocabulary Object Detection by Novel-Class Feature Perception Enhancement

3.1 Pretraining by Image Masking

3.2 IDF Feature Extraction Network

4.1 Setup

4.2 Comparison with State-of-the-Art Approaches

4.3 Ablation Study

References.

PLOD-YOLO: Premium Lightweight Object Detection for Autonomous Following Robot.

Notes:

Includes bibliographical references and index.

ISBN:

981-9755-91-3