Intelligence Science IV: 5th IFIP TC 12 International Conference, ICIS 2022, Xi'an, China, October 28–31, 2022, Proceedings (IFIP Advances in Information and Communication Technology, 659)

This book constitutes the refereed proceedings of the 5th International Conference on Intelligence Science, ICIS 2022, held in Xi'an, China, in August 2022. The 41 full and 5 short papers presented in this book were carefully reviewed and selected from 85 submissions. They were organized in topical sections as follows: Brain cognition; machine learning; data intelligence; language cognition; remote sensing images; perceptual intelligence; wireless sensor; and medical artificial intelligence. Preface Organization Abstracts of Keynote and Invited Talks Tactile Situations: A Basis for Manual Intelligence and Learning Brain-like Perception and Cognition: Challenges and Thinking Dealing with Concept Drifts in Data Streams A Novel Bionic Imaging and Its Intelligent Processing Skill Learning in Dynamic Scene for Robot Operations Emerging Artificial Intelligence Technologies in Healthcare Memory Cognition Contents Brain Cognition Mouse-Brain Topology Improved Evolutionary Neural Network for Efficient Reinforcement Learning 1 Introduction 2 Related Works 3 Methods 3.1 The Allen Mouse Brain Atlas 3.2 The Clustered Hierarchical Circuits 3.3 The Neuron Model 3.4 Coping the Biological Circuits to Artificial Ones 3.5 The Network Learning 4 Experiments 4.1 The Clustered Brain Regions 4.2 The Network Topology from Biological Mouse Brain 4.3 Results with Circuit-46 and Random Networks 4.4 Result Comparison with Different Algorithms 5 Discussion References DNM-SNN: Spiking Neural Network Based on Dual Network Model 1 Introduction 2 Methods 2.1 Traditional SNN Supervised Learning Algorithm Framework and Its Limitations 2.2 Proposed Dual-Model Spike Network Supervised Learning Algorithm 2.3 Proposed Multi-channel Mix Module Prediction Method 2.4 The Chosen Network Model 2.5 Selection of Spiking Neurons 3 Experimental Results 3.1 Single- and Dual-Model Resnet11 Performance on the CIFAR-10 Dataset 3.2 Related Work Comparison 4 Conclusion References A Memetic Algorithm Based on Adaptive Simulated Annealing for Community Detection 1 Introduction 2 Background 2.1 Modularity 2.2 Normalized Mutual Information 3 Description of MA-ASA 3.1 Segmented Label Propagation 3.2 Selection and Crossover Operation 3.3 Mutation Operation 3.4 Improved Simulated Annealing 3.5 Framework of MA-ASA 4 Experiments and Analysis 4.1 Experimental Settings 4.2 Experimental Results and Analysis 5 Conclusion References The Model of an Explanation of Self and Self-awareness Based on Need Evolution 1 Background and Significance 2 The Nature and Needs of Life 2.1 The Nature and Representation of the Self 2.2 The Primary Needs and Principle of Life 3 Evolution and Representation of the Needs of Life 3.1 Needs Representation and Original Self-evolution in Single-Celled and Complex Organisms 3.2 Representation Needs and Self-awareness of Human 4 Self-model Based on the Evolution of Needs 4.1 Iterative Model of Needs Evolution 4.2 Evolutionary Model of the Self 5 Dicussion and Conclusion References Spiking Neuron Network Based on VTEAM Memristor and MOSFET-LIF Neuron 1 Introduction 2 Proposed Method 2.1 Leaky Integrate-and-Fire Model 2.2 Design of LIF Circuit 2.3 Correspondence Between Network and Circuit 2.4 Processing of the DVS128 Gesture Dataset 2.5 Network Formulation 3 Performance Analysis and Discussion 4 Conclusion References Machine Learning A Deception Jamming Discrimination Method Based on Semi-supervised Learning with Generative Adversarial Networks 1 Introduction 2 Signal Model 2.1 The Construction of a Multistatic Radar System Model 2.2 Generation of Echo Data 3 The Discrimination Network Based on SGAN 4 Simulation 4.1 Simulation Analysis 4.2 Simulation Results with Different PRI 4.3 The Comparison of Different Discrimination Methods 5 Conclusion References Fast Node Selection of Networked Radar Based on Transfer Reinforcement Learning 1 Introduction 2 Related Work 2.1 Radar Node Selection 2.2 Reinforcement Learning 2.3 Transfer Learning 3 Methodology 3.1 Revisiting of Monte Carlo Tree 3.2 The Lower Bound of Cramero (CLRB) 3.3 Selection Flow 3.4 Variable-Number Node Search 3.5 Transfer Reinforcement Learning 4 Experiments and Analysis 5 Conclusion References Weakly Supervised Liver Tumor Segmentation Based on Anchor Box and Adversarial Complementary Learning 1 Introduction 2 Approach 2.1 Anchor Boxes Generation 2.2 Adversarial Complementary Learning 2.3 Application 2.4 Pseudo Mask Generation 3 Experiments 3.1 Datasets and Evaluated Metric 3.2 Classification Network and Hyperparameter Settings 3.3 Segmentation Network and Test Results 4 Conclusions References Weakly Supervised Whole Cardiac Segmentation via Attentional CNN 1 Introduction 2 Method 2.1 Pseudo Masks 2.2 Deep U-Net Network 2.3 Improved Weighted Cross-Entropy Loss 3 Experimental and Results 3.1 Datasets and Implementation Details 3.2 Patch Selection 3.3 Experimental Results 3.4 Ablation Experiments 3.5 Generality Experiments 4 Conclusion References Noisy Label Learning in Deep Learning 1 Introduction 2 Preliminary Knowledge 2.1 Noisy Labels in Deep Learning 2.2 Noisy Label Dataset and Noisy Label Types 2.3 Analysis the Problems in Noisy Label Learning 3 Existing Methods of Noisy Label Learning 3.1 Full-Equal-Using Method 3.2 Clean-Based Method 3.3 Full-Differ-Using Method 4 Problems in Existing Methods 4.1 Difference Between Synthetic Dataset and the Actual Dataset 4.2 Problems with Existing Methods 4.3 Possible Solutions 5 Conclusion References Accelerating Deep Convolutional Neural Network Inference Based on OpenCL 1 Introduction 2 Related Work 3 Design, Implementation and Optimization of CNN on OpenCL 3.1 Parallel Strategy for Convolution Layer 3.2 Parallel Strategy for Other Layers 3.3 Kernel Fusion and Increasing Global Task 4 Experiment and Evaluations 4.1 Experimental Environment 4.2 Performance Comparison of Depthwise Convolution Operations 4.3 Comparison of Parallel DCNN Inference Performance 4.4 Performance Comparison of Different Hardware Environments 5 Conclusions References A Simple Approach to the Multiple Source Identification of Information Diffusion 1 Introduction 2 Related Works and Motivations 2.1 Related Methods 2.2 Motivations 3 Preliminaries and Problem Formulation 3.1 Susceptible-Infected (SI) Model 3.2 Problem Formulation 4 KST Method 4.1 Analysis 4.2 KST Method 5 KST-Improved Method 6 Evaluation 6.1 Experiments Settings 6.2 Accuracy of Identifying Sources 7 Conclusion References Data Intelligence A Directed Search Many Objective Optimization Algorithm Embodied with Kernel Clustering Strategy 1 Introduction 2 The Proposed Method 2.1 Directed Search Sampling and Guiding Solutions 2.2 Environmental Selection 3 Experimental Results and Analysis 4 Conclusion References A Two-Branch Neural Network Based on Superpixel Segmentation and Auxiliary Samples 1 Introduction 2 Proposed Method 2.1 Selection of Auxiliary Samples 2.2 The Structure of TBN-SPAS 3 Implementation Process of TBN-MERS 4 Experiment and Analysis 4.1 Experimental Settings 4.2 The Role of Auxiliary Branch 4.3 Comparison with Existing Methods 5 Conclusions References Augmentation Based Synthetic Sampling and Ensemble Techniques for Imbalanced Data Classification 1 Introduction 2 Augmentation Based Synthetic Sampling Method 2.1 Data Augmentation (DA) 2.2 Notations 2.3 Proposed Method 3 Experiment Settings and Result Analysis 3.1 Datasets 3.2 Evaluation Metric 3.3 Experimental Results 4 Integration of Augmentation Based Synthetic Sampling Method and Ensemble Techniques 5 Conclusion References Language Cognition BA-GAN: Bidirectional Attention Generation Adversarial Network for Text-to-Image Synthesis 1 Introduction 2 Related Work 3 Our Model 3.1 Text Encoder and Image Encoder 3.2 Multi-stage Generative Adversarial Networks 4 Experiments 5 Conclusion References Personalized Recommendation Using Extreme Individual Guided and Adaptive Strategies 1 Introduction 2 Background 2.1 Definition of Recommendation Problem 2.2 Multi-objective Optimization Problem 2.3 Probs 3 Proposed Algorithm 3.1 Framework of MOEA-EIMA 3.2 Individual Encoding and Initialization 3.3 The Two Objectives 3.4 Genetic Operators 4 Experiments and Analysis 4.1 Experiment Settings 4.2 Experimental Results 5 Conclusions References Improved Transformer-Based Implicit Latent GAN with Multi-headed Self-attention for Unconditional Text Generation 1 Introduction 1.1 Generative Adversarial Network (GAN) for Unconditional Text Generation 1.2 Research Objective and Content 2 Related Works 3 Model Architecture 3.1 Overall Framework 3.2 Multi-headed Self Attention Based Generator 3.3 Training Details 4 Experiments 4.1 Evaluation Metrics 4.2 Microsoft COCO: Common Objects in Context 4.3 Ablation Experiment 5 Conclusion and Future Work References Learning a Typhoon Bayesian Network Structure from Natural Language Reports 1 Introduction 2 Related Works 3 The Framework of Learning Typhoon Bayesian Network Structures 3.1 State Extraction Model 3.2 Standardize State Information 3.3 Causal Relationship Extraction 3.4 Generate Typhoon Bayesian Network 4 Experimental Results 5 Discussions and Conclusions References What Is Information? An Interpretation Based on the Theory of Modern Complexity Science 1 Introduction 2 Material Hierarchies and Complexity Systems 3 How Information Comes into the View of Scientists 3.1 The Reduction of Three Types of Determinism 3.2 Interpretation of Information on Biological Genetics 3.3 Information and Communication Connections Are Established between Different and Across Layers of the Organism 4 The Interaction between Layers in Complexity System and Its Informatization 4.1 The Change of Causal Ideas 4.2 Information Entity 4.3 The Role of the Hierarchical Interactions 5 Short Conclusion References Remote Sensing Images Deep Siamese Network with Contextual Transformer for Remote Sensing Images Change Detection 1 Introduction 2 Methodology 2.1 Feature Extractor 2.2 Transformer Module 2.3 Prediction Module 2.4 Details of Loss Function 3 Experiment 3.1 Dataset 3.2 Implementations 3.3 Results and Analyses 4 Conclusion References GSoP Based Siamese Feature Fusion Network for Remote Sensing Image Change Detection 1 Introduction 2 Methodology 2.1 GSoP Module 2.2 Information Fusion Based on GSoP 2.3 Siamese Feature Fusion Network with GSoP for Change Detection 3 Experiments and Analysis 3.1 Dataset Introduction and Experiments Setting 3.2 Experimental Results and Analysis 4 Conclusion References PolSF: PolSAR Image Datasets on San Francisco 1 Introduction 2 PolSF 2.1 PolSF-AIRSAR 2.2 PolSF-ALOS2 2.3 PolSF-GF3 2.4 PolSF-RISAT 2.5 PolSF-RS2 3 Conclusions References RSMatch: Semi-supervised Learning with Adaptive Category-Related Pseudo Labeling for Remote Sensing Scene Classification 1 Introduction 2 Proposed Method 2.1 Preliminary Knowledge 2.2 Adaptive Class-Related Pseudo Labeling 3 Experiments 3.1 Experimental Settings 3.2 Classification Results 4 Conclusion References Visual Question Answering of Remote Sensing Image Based on Attention Mechanism 1 Introduction 2 Related Work 2.1 Co-attention 3 Methods 3.1 Hybrid Moudlar Co-attention 3.2 Cross-Modal Fusion of Global and Local Information 4 Experiments 4.1 Dataset 4.2 Experimental Setup and Hyperparameters 4.3 Experimental Results and Analysis 5 Conclusion References Multi-scale Spatial Aggregation Network for Remote Sensing Image Segmentation 1 Introduction 2 Related Work 2.1 Semantic Segmentation of Natural Scene Image 2.2 Semantic Segmentation of Remote Sensing Image 3 Proposed Method 3.1 The Architecture of MSAN 3.2 Densely Connected Structure 3.3 Multi-scale Information Fusion Module 3.4 Spatial Path 3.5 Smoothing Algorithm 4 Experimental Results and Analysis 4.1 Dataset 4.2 Experimental Setting 4.3 Experimental Results 4.4 Ablation Experiments 5 Conclusions References Deep Complex Convolutional Neural Networks for Remote Sensing Image Classification 1 Introduction 2 Related Work 2.1 Convolutional Neural Network 2.2 ENet 3 Methodology 3.1 Complex Convolution Network 3.2 Network Parameters Analysis 3.3 Complex Batch Norm 3.4 Complex Activate Function 4 Experiment 5 Conclusion References Perceptual Intelligence Dual Siamese Channel Attention Networks for Visual Object Tracking 1 Introduction 2 Related Work 2.1 Siamese Based Trackers 2.2 Attention Mechanism 3 Method 3.1 Siamese-Based Trackers 3.2 Convolutional SE Networks in Backbone 3.3 Global Channel Enhancement Module 4 Experiments 4.1 Implementation Details 4.2 Comparisons with the State-of-the-Art 4.3 Ablation Study 5 Conclusion References Motion-Aligned and Hardness-Aware Dynamic Update Network for Weakly-Supervised Vehicle Detection in Satellite Videos 1 Introduction 2 Proposed Method 2.1 Overview 2.2 Motion-Aligned Initialization 2.3 Motion-Aligned Initialization 2.4 Online Pseudo Label Update Scheme 3 Experimental Results and Analysis 3.1 Datasets 3.2 Experimental Setups 3.3 Detection Results of Different Methods 3.4 Ablation Experiments 4 Conclusion References A Multi-level Mixed Perception Network for Hyperspectral Image Classification 1 Introduction 2 Methodology 3 Experiments Results and Analysis 3.1 Parameters Analysis and Ablation Study 3.2 Comparison with State-of-the-Art Methods 4 Conclusion References A Lightweight SAR Ship Detection Network Based on Superpixel Statistical Modeling 1 Introduction 2 Superpixel-Based Composite Gamma Distribution Modeling 2.1 Superpixel 2.2 Composite Gamma Distribution Based on Superpixel 3 Lightweight Ship Detection Network 3.1 ShuffleNeck Module 3.2 Shuffle-X(SX) Module 3.3 YOLO-SX Algorithm Description 4 Experiment 4.1 Experiment Dataset and Details 4.2 Ablation Experiment 4.3 Comparison Experiment 5 Conclusion References Background Augmentation with Transformer-Based Autoencoder for Hyperspectral Anomaly Detection 1 Introduction 2 Methodology 2.1 Typical Background Samples Selection 2.2 Transformer-Based Autoencoder 2.3 RX Anomaly Detection Algorithm 3 Experiments and Results 3.1 Datasets and Implementation 3.2 Experimental Results 4 Conclusion References Point Cloud Registration Based on Global and Local Feature Fusion 1 Introduction 2 Related Work 2.1 Traditional Registration Method 2.2 Deep-Learning Registration Method 2.3 GLF-DQNet 3 Method 3.1 Problem Statement 3.2 SGLFNet Network 3.3 Loss Function 4 Experiments 4.1 Experiments on GLFNet 4.2 Experiments on Local Registration 4.3 Experiments on Global Registration 5 Conclusion References Gaussian Balanced Sampling for End-to-End Pedestrian Detector 1 Introduction 2 Approach 2.1 Gaussian Negative Sampling 2.2 PD Loss 2.3 Non-target Response Suppression 3 Experiments 3.1 Performance on CrowdHuman Dataset 3.2 Ablation Study 4 Conclusion References Combining Spatial-Spectral Features for Hyperspectral Image Few-Shot Classification 1 Introduction 2 Proposed Approach 2.1 Feature Extraction 2.2 Source and Target Few-Shot Learning 2.3 Domain Alignment 3 Experimental Results 3.1 Experimental Setting and Performance 4 Conclusions References SAR Scene Classification Based on Self-supervised Jigsaw Puzzles 1 Introduction 2 Methodology 2.1 Jigsaw Puzzle Problem 2.2 Network Structure 2.3 Network Training 2.4 Shortcuts Prevention 2.5 Downstream Tasks 3 Experiments 3.1 Dataset and Experiment Settings 3.2 Upstream Task Results 3.3 Downstream Task Results 4 Conclusions References YOLO-Head: An Input Adaptive Neural Network Preprocessor 1 Introduction 2 Method 2.1 YOLO-Head 2.2 Pipeline 3 Experiments 3.1 Settings 3.2 Evaluation for YOLO-Head 3.3 Application in Object Detector 4 Conclusion References SR-YOLO: Small Objects Detection Based on Super Resolution 1 Introduction 2 Related Work 2.1 Object Detection 2.2 Image Super Resolution 3 Method 3.1 Image Super Resolution Based on Generative Adversarial Networks 3.2 Feature Extraction Backbone Considering Edge Information 4 Experiment 4.1 Datasets 4.2 Experimental Environment and Training Settings 4.3 Evaluation Indicators 4.4 Analysis of Experimental Results 5 Conclusion References Multi Recursive Residual Dense Attention GAN for Perceptual Image Super Resolution 1 Introduction 2 Background 3 Proposed Methods 3.1 Our Multi-recursive Residual Dense Network in the Generator 3.2 Our CA Network in the Discriminator 3.3 Loss Functions 4 Experiments and Analysis 4.1 Implementation Details 4.2 Experimental Data 4.3 Ablation Studies 4.4 Experimental Results 5 Conclusion and Further Work References Relay-UNet: Reduce Semantic Gap for Glomerular Image Segmentation 1 Introduction 2 Related Work 2.1 Glomerular Image Segmentation 2.2 UNet Structure 3 Methodology 3.1 Relay-UNet Structure 3.2 Improved Method 4 Experiment and Result Analysis 4.1 Dataset 4.2 Evaluating Metrics 4.3 Training Process 5 Conclusion References MobiCFNet: A Lightweight Model for Cattle Face Recognition in Nature 1 Introduction 2 Related Work 3 Fine-Grained Face Individual Recognition Dataset 4 Proposed Method 4.1 CNN and Lightweight CNN 4.2 Large Margin Cosine Loss 4.3 MobiCFNet 5 Experiment 6 Conclusion and Future Scopes References Molecular Activity Prediction Based on Graph Attention Network 1 Introduction 1.1 Graph Neural Networks 1.2 Attention Mechanism 1.3 Motivation and Contribution 1.4 Organization 2 Method 2.1 The Overall Framework 3 Experiment 3.1 Experimental Setting 3.2 Experimental Result 4 Conclusion References Tracking Multi-objects with Anchor-Free Siamese Network 1 Introduction 2 SiamBAN-MOT 2.1 Anchor-Free Network with Detector and S-FPN 2.2 Siamese Tracker for Multi-object Tracking. 2.3 Ground-Truth and Loss 3 Experiment 3.1 Datasets and Metrics 3.2 Implementation Details 3.3 Experiment Results on MOT17 4 Conclusion References Feature Learning and Change Feature Classification Based on Variational Auto-encoder for SAR Change Detection 1 Introduction 2 The Proposed Method 2.1 Preprocessing and FCM 2.2 SVAE for Feature Learning and Classification 3 Experimental Settings and Results Analysis 3.1 Data Description 3.2 General Information 3.3 Analysis of the Representation Ability 3.4 Results Comparison with Other Methods on Ottawa Data Set 3.5 Results Comparison with Other Methods on San Francisco Set 4 Conclusion References A Simple Structure for Building a Robust Model 1 Introduction 2 Related Work 2.1 Adversarial Attacks 2.2 Adversarial Defensive 3 Methods 3.1 Structure Design 3.2 Sampling Strategy 4 Experiments 5 Conclusion References Wireless Sensor An Adaptive Spatial Network for UAV Image Real-Time Semantic Segmentation 1 Introduction 1.1 Related Work 1.2 Motivation 1.3 Contributions 2 Proposed Approach 2.1 Structure of ASRnet 2.2 Adaptive Local Feature Descriptor 3 Experiments 4 Conclusion References Medical Artificial Intelligence Knowledge Learning Without Forgetting for the Detection of Alzheimer's Disease 1 Introduction 2 Method 2.1 Transfer Learning for AD Detection 2.2 Knowledge Learning without Forgetting 2.3 Transfer Based on Improved Contrastive Loss 3 Experiments 3.1 Datasets 3.2 Experimental Setting 3.3 Experimental Results 4 Conclusion References CA-ConvNeXt: Coordinate Attention on ConvNeXt for Early Alzheimer's Disease Classification 1 Introduction 2 Method 2.1 ConvNeXt 2.2 Coordinate Attention 3 Experiments and Results 3.1 Data Pretreatment 3.2 Experimental Details 3.3 Experimental Result 4 Conclusions References Data Augmentation Method on Pine Wilt Disease Recognition 1 Introduction 2 Augmentations 3 Materials and Methods 4 Results 5 Discussion and Conclusion References Correction to: Weakly Supervised Whole Cardiac Segmentation via Attentional CNN Correction to: Chapter “Weakly Supervised Whole Cardiac Segmentation via Attentional CNN” in: Z. Shi et al. (Eds.): Intelligence Science IV, IFIP AICT 659, https://doi.org/10.1007/978-3-031-14903-0_9 Author Index