Pattern Recognition

Name: Pattern Recognition | 27th International Conference, ICPR 2024, Kolkata, India, December 1-5, 2024, Proceedings, Part VII
Brand: Springer
Price: 78.1 EUR
Availability: OnlineOnly

27th International Conference, ICPR 2024, Kolkata, India, December 1-5, 2024, Proceedings, Part VII

Apostolos Antonacopoulos Subhasis Chaudhuri Rama Chellappa Cheng-Lin Liu Saumik Bhattacharya Umapada Pal(Herausgeber*in)

Springer (Verlag)

Erschienen am 3. Dezember 2024

XXXVIII, 490 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-031-78183-4 (ISBN)

78,10 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Weitere Ausgaben

Inhalt

Intro
President's Address
Preface
Organization
Contents - Part VII
Graph Matching Networks Meet Optimum-Path Forest: How to Prune Ensembles Efficiently
1 Introduction
2 Related Works
3 Theoretical Background
3.1 Optimum-Path Forest
3.2 Graph Matching Network
4 Proposed Approach
5 Methodology
5.1 Dataset
5.2 Experimental Setup
6 Results and Discussions
7 Conclusions
References
Understanding the Influence of Extremely High-Degree Nodes on Graph Anomaly Detection
1 Introduction
2 Related Works
3 Data Collection and Properties
4 Exploring the Influence of Extremely High-Degree Node
4.1 Definition of Extremely High-Degree Node
4.2 Experimental Settings
4.3 Influence of SN on GNN-Based and Non-GNN-Based Models
4.4 Impact of SN on Unsupervised and Supervised GADs
4.5 Computational Cost
4.6 Over-Smoothing
5 Method and Experiments
5.1 SNGNN
5.2 Experiments
5.3 Ablation Study
6 Conclusion and Limitation
References
Spatio-Temporal Heterogeneous Graph Neural Network With Multi-view Learning For Traffic Prediction
1 Introduction
2 Related Work
2.1 Traffic Forecasting
2.2 Graph Neural Networks
2.3 Dynamic Graph Learning
3 Problem Definition
4 Methodology
4.1 Heterogeneous Temporal Convolution Module
4.2 Dynamic Graph Learning Module
4.3 Dynamic Graph Convolution Module
4.4 Multi-view Fusion Module
4.5 Output Module
5 Experimental Studies
5.1 Datasets and Evaluation Metrics
5.2 Experimental Settings
5.3 Baselines
5.4 Comparison Results
5.5 Ablation Study
5.6 Effects of Multi-view and Dynamic Graph Learning
6 Conclusion
References
BotSCL: Heterophily-Aware Social Bot Detection with Supervised Contrastive Learning
1 Introduction
2 Related Work
2.1 Graph-Based Social Bot Detection
2.2 GNNs for Graphs with Heterophily
2.3 Contrastive Learning
3 Methodology
3.1 Graph Augmentation
3.2 Aggregation Strategy
3.3 Supervised Contrastive Optimization
4 Experiments
4.1 Experiment Setup
4.2 Heterophily Evidence and Influence
4.3 Performance Comparison
4.4 Ablation Study
4.5 Sensitive Analysis
4.6 Visualization
5 Conclusion
References
SimDrop: Towards Deep Graph Convolutional Networks
1 Introducation
2 Related Work
3 Preliminaries
4 Method
5 Experiment
5.1 Experiment Setting
5.2 Experimental Results and Analysis
6 Conclusion
References
A Quantum-inspired Approach to Estimate Optimum-Path Forest Prototypes based on the Traveling Salesman Problem
1 Introduction
2 Theoretical Background
2.1 Optimum-Path Forest
2.2 Quantum Machine Learning
3 Quantum-inspired Prototype Computation
4 Methodology
4.1 Datasets
4.2 Experimental Setup
5 Experiments and Results
5.1 Convergence Analysis
5.2 Discussion
6 Conclusions
6.1 Challenges
6.2 Future Works
References
Face to Cartoon Incremental Super-Resolution Using Knowledge Distillation
1 Introduction
2 Related Work
2.1 Incremental Learning and Knowledge Distillation
2.2 Face Super-Resolution
3 Proposed Methodology
3.1 Problem Description
3.2 Knowledge Distillation
3.3 Edge Block
3.4 Generator Architecture
3.5 Discriminator Architecture
3.6 Objective Function
4 Experimental Settings
5 Experimental Results and Discussion
5.1 Quantitative Results
5.2 Qualitative Results
5.3 Ablation Study on Loss Hyperparameters
5.4 Cross-Dataset Analysis
5.5 Comparsion with Joint Training Approach
5.6 Performance on Extended Network
6 Conclusion
References
Copula Entropy Based Causal Network Discovery from Non-stationary Time Series
1 Introduction
2 Related Work
3 Preliminary Knowledge
3.1 Mutual Information and Transfer Entropy
3.2 Conditional Transfer Entropy Estimation
3.3 Definitions and Assumptions in Non-stationary Time Series Causal Discovery
4 Algorithm Introduction
4.1 Causal Structure Learning of the Non-stationary Time Series
4.2 Estimation of the Conditional Transfer Entropy
4.3 Time Complexity Analysis
5 Experimental Results and Analysis
5.1 Experiment 1
5.2 Experiment 2
5.3 Experiment 3
5.4 Experimental Analysis
6 Real Data Experiment
7 Conclusion
References
DSparsE: Dynamic Sparse Embedding for Knowledge Graph Completion
1 Introduction
2 Background and Related Works
3 DSparsE for Link Prediction
3.1 Dynamic Layer
3.2 Relation-Aware Layer
3.3 Projection Layer
3.4 Residual Layer
3.5 Sparse Structure of MLP
4 Experiments and Analysis
4.1 Datasets and Evaluation Settings
4.2 Prediction Performance
4.3 Ablation Studies and Further Experiments
5 Conclusion
References
Interpreting Convolutional Neural Network Decision via Pixel-Wise Interaction Hierarchy Graph
1 Introduction
2 Related Work
3 Method
3.1 Interaction Value
3.2 Interaction Hierarchy Graph
3.3 Filter Contribution
4 Experiments
4.1 Experimental Settings
4.2 Visual Knowledge of Filters
4.3 Quantitative Evaluation
4.4 Faithfulness Evaluation
5 Conclusion
References
Denoising Optimization-Based Counterfactual Explanations for Time Series Classification
1 Introduction
2 Preliminaries
2.1 Problem Formalism
2.2 Wachter's Method
2.3 Contrastive Explanation Method
2.4 Counterfactual Explanations Guided by Prototypes
2.5 TimeX
2.6 Discrete Fourier Transform
3 Proposed Approach
3.1 Motivation
3.2 Low-Pass Filtering
3.3 Frequency Clipping
3.4 Time Complexity
4 Experimental Setup
4.1 Black-Box Classification Models
4.2 Datasets
4.3 Implementation Details
5 Experimental Results
5.1 Plausibility
5.2 Proximity
5.3 Visual Plausibility
5.4 PCA
5.5 Sensitivity Analysis
6 Conclusion
References
Improving Adaptive Runoff Forecasts in Data-Scarce Watersheds Through Personalized Federated Learning
1 Introduction
2 Related Work
3 Methodology
3.1 Problem Formulation
3.2 FedHydroDSW: Federated Runoff Forecast for Data-Scarce Watershed
4 Empirical Analysis
4.1 Dataset and Preprocessing
4.2 Design of the Experimental Study
4.3 Experimental Result and Discussion
4.4 Ablation Study
5 Discussion
6 Conclusion
References
Stagger-Cache MITM: A Privacy-Preserving Hierarchical Model Aggregation Framework
1 Introduction
2 Related Work
3 Contributions of This Paper
4 Problem Setup
4.1 Mathematical Notation
4.2 Performance Criteria
5 Multi-tier Inference Frameworks
5.1 Framework A: Bottom-Up
5.2 Computation Methodology
5.3 Framework B: Tier-Caching
5.4 Computation Methodology
6 Framework of Meet-In-The-Middle (MITM) Staggered Caching
6.1 Computation Methodology
6.2 Benefits and Drawbacks
7 Experimental Results
7.1 Use Case: Crop Disease Forecasting
7.2 Results: Inference Latency
7.3 Results: Accuracy
7.4 Results: Memory Used
7.5 Results: Data Transmitted over Network
7.6 Experiments Discussion
8 Conclusion
References
ViT2 - Pre-training Vision Transformers for Visual Times Series Forecasting
1 Introduction
2 Related Work
3 Background
3.1 Gramian Angular Fields
3.2 Vision Transformer
4 Methodology
4.1 Module 1 - Data Preprocessing
4.2 Module 2 - Visualizing Time Series
4.3 Module 3 - Probabilistic Forecasting ViT
4.4 Module 4 - Transfer Learning & Fine-Tuning
5 Experiments and Discussion
5.1 Datasets
5.2 Experimental Setup
5.3 Hyper-parameter Tuning
5.4 Results
6 Conclusion & Future Work
References
waLLMartCache: A Distributed, Multi-tenant and Enhanced Semantic Caching System for LLMs
1 Introduction
2 Related Work
3 GPTCache in a Nutshell
3.1 Adapter
3.2 Pre-processor
3.3 Embedding Generator
3.4 Cache Manager
3.5 Similarity Evaluator
3.6 Post-processor
4 waLLMartCache: An Enhanced Cache for LLMs
4.1 Incorporating Redis as a Database
4.2 Designing a Distributed Cache for LLMs
4.3 Integrating Multi-tenancy
4.4 Improved Decision Engine for Caching
4.5 Pre-loading (non-volatile) FAQs into the Cache
4.6 Ablation Study
5 Conclusion
References
ReeSPOT: Reeb Graph Models Semantic Patterns of Normalcy in Human Trajectories
1 Introduction
2 Methodology
2.1 Previous work on Reeb graphs
2.2 Reeb graph models agent pattern of normalcy
2.3 Construction of Reeb graphs and analysis of time complexity
3 Experimentation/Case Study
3.1 Data generation
3.2 Definition of anomalous behavior
3.3 Reeb Graph Generation
3.4 Analysis and interpretation of scenarios using Reeb graphs
3.5 Reeb graph iteratively detects anomalous behavior of an agent
3.6 Quantifying the distance between Reeb graphs
3.7 Scalability with Reeb Graphs
4 Discussion and Future Work
References
Label Disambiguation-Based Feature Selection for Partial Multi-label Learning
1 Introduction
2 Related Work
3 Proposed Approach
3.1 Label Disambiguation Using Granular Ball
3.2 Feature Selection Using Labeling Confidence
4 Experiments
4.1 Datasets
4.2 Baselines
4.3 Experimental Results
4.4 Ablation Study
5 Conclusion
References
Neural Encoding of Odors: Translating Odors into Unique Digital Representation with EEG Signals
1 Introduction
2 Related Work
3 Methodology
3.1 Data Collection
3.2 Data Pre-processing
3.3 Encoder Network Architecture
3.4 Unique Digital Representation
4 Result and Discussion
4.1 Encoding Representations Analysis
4.2 Odor Matrix Analysis
5 Conclusion
References
Robust Feature Space Organization with Distillation for Few-Shot Object Detection
1 Introduction
2 Related Works
2.1 Few-Shot Object Detection
2.2 Contrastive Learning
3 Method
3.1 Problem Formulation
3.2 Approach
4 Implementation Details and Experimental Results
4.1 Benchmark Datasets and Evaluation Metrics
4.2 Implementation Details
4.3 Results
4.4 Ablation Studies
4.5 Qualitative Results
5 Conclusion
References
Image Domain Translation for Few-Shot Learning
1 Introduction
2 Related Work
3 Methodology
3.1 Background
3.2 Overall Framework of SCTN
4 Experiments
4.1 Datasets
4.2 Implementation Details
4.3 Results
4.4 Ablation Study
4.5 Visualization Analysis
5 Conclusion
References
Towards Adversarial Robustness and Reducing Uncertainty Bias through Expert Regularized Pseudo-Bidirectional Alignment in Transductive Zero Shot Learning
1 Introduction
2 Related Works
2.1 Zero-Shot Learning
2.2 Graph Neural Networks in ZSL
3 Limitations of Bidirectional Alignment
4 Methodology
4.1 Problem Formulation
4.2 Overall Outline
4.3 Training the classifier
4.4 Theoretical Perspective
5 Experiments
5.1 Performance Comparison
5.2 Ablation studies
6 Conclusions
References
Zero-Shot Underwater Gesture Recognition
1 Introduction
2 Related Works
2.1 Hand Gesture Recognition (HGR)
2.2 Zero-Shot Learning (ZSL)
2.3 Underwater Diver Gesture Recognition
3 Methodology
3.1 Problem Definition
3.2 System Overview
3.3 Gated Cross-Attention Transformer (GCAT)
3.4 Transformed-Feature Generating Network
3.5 Training and Inference
4 Experiments
4.1 Dataset
4.2 Zero-Shot Splits and Evaluation Protocols
4.3 Implementation Details
4.4 Zero-Shot Results
4.5 Ablation Studies
5 Conclusion
References
Optic Atrophy Classification from Fundus Images with Few-Shot Learning
1 Introduction
1.1 Motivation
1.2 Objectives
1.3 Literature Survey
1.4 Research Issues
1.5 Contributions
2 Dataset
2.1 Data Augmentation
2.2 Dataset Split
3 Network Architecture
4 Hyper-parameter Optimization
4.1 Hyper-parameter Space
4.2 Baseline Model Training
4.3 Training the Siamese Network
4.4 Embedding Space Visualization
5 Results
5.1 Distance Metric Evaluation
5.2 Comparison With Existing Works
5.3 Ablation Study
6 Conclusion
References
Recognition of Online Handwritten Chinese Texts in Any Writing Direction via Stroke Classification Based Over-Segmentation
1 Introduction
2 Related Work
2.1 Text Over-Segmentation
2.2 Text Recognitions in Any Writing Direction
3 Methodology
3.1 Overview
3.2 Classification of Text Line Inclination Style
3.3 Stroke Classification Model Based on BiLSTM
4 Experiments
4.1 Dataset
4.2 Performance Metrics
4.3 Implementation Details
4.4 Ablation Experiment
4.5 Comparison with the State-of-the-Art Methods in the Horizontal Text Recognition Task
4.6 Text Recognition Experiment in Any Writing Direction
4.7 Generalization Experiment
4.8 Further Visualization and Analysis
5 Conclusions
References
ProFONet: Prototypical Feature Space Optimized Network for Few Shot Classification
1 Introduction
2 ProFONet: Intuition
3 ProFONet: Detailed Description
3.1 Method
3.2 Implementation Details
4 Results
4.1 Evaluation on CUB
4.2 Evaluation on GI-Findings
5 Analysis
5.1 Effectiveness of VIC Based Feature Optimization
5.2 Ablation Study with Different Backbones
5.3 Ablation Study with Different Distance Functions
5.4 Effects of Deep Supervision (DS) and VIC Injected ProtoLoss
6 Conclusion
References
Few-Shot Copycat: Improving Performance of Black-Box Attack with Random Natural Images and Few Examples of Problem Domain
1 Introduction
2 Related Works
3 Few-Shot Copycat
3.1 Few-Shot Fake Dataset Generation
4 Experimental Methodology
4.1 Baselines
4.2 Use Case Problems
4.3 Metrics
4.4 Experiment
4.5 General Setup
5 Experimental Results
5.1 Discussion
6 Conclusion
References
Learning Using Generated Privileged Information by Text-to-Image Diffusion Models
1 Introduction
2 Related Work
3 Method
4 Experiments
4.1 Data Sets
4.2 Experimental Setup
4.3 Results
5 Conclusion
References
Deep Hardware Modality Fusion for Image Segmentation
1 Introduction
2 Related Work
3 Method
3.1 Hardware Modality Fusion
3.2 Segmentation Model
3.3 Training and Inference
4 Experiments
4.1 Datasets
4.2 Training Details
4.3 Comparison with Multimodal Methods in Digital Domain
4.4 Discussion on Frame Coding and Pixel Coding
4.5 Ablation Study
5 Conclusion
References
Temporal Insight Enhancement: Mitigating Temporal Hallucination in Video Understanding by Multimodal Large Language Models
1 Introduction
2 Research Background
2.1 MLLMs
2.2 Hallucination in MLLMs
2.3 Hallucination Correction
3 Method
3.1 Two Tasks for Temporal Hallucination Evaluation
3.2 Event Temporal Hallucination Correction
3.3 Claim Integration for Response Correction.
4 Experiment
4.1 Experimental Setting
4.2 Temporal Hallucination Evaluation and Correction Result
4.3 Ablation Experiment for External Tools
4.4 Different Prompt Instruction for Response
5 Discussion
5.1 Advantages Over Fine-Tuning MLLM Solution
5.2 Performance Impact of LLMs and External Tools
5.3 Choice of Evaluation Metrics
5.4 Scalability and Generalization
6 Conclusion
References
MC-DBN: A Deep Belief Network-Based Model for Modality Completion
1 Introduction
2 Related Work
2.1 Multi-modal Data Integration
2.2 Stock Market Forecasting and Heart Rate Monitoring
2.3 Methods for Handling Missing Values
3 Methodology
3.1 RBM-Based Latent Representation Learning
3.2 Modal Completion Encoder-Decoder Framework
3.3 Attention Fusion Module
3.4 Multiple Loss Function Design
3.5 Model Training and Evaluation
4 Experiment
4.1 Data Preparation
4.2 Comparative Experiment
4.3 Ablation Experiment
5 Conclusion
References
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken