Advances in Knowledge Discovery and Data Mining : 28th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2024, Taipei, Taiwan, May 7–10, 2024, Proceedings, Part V / edited by De-Nian Yang, Xing Xie, Vincent S. Tseng, Jian Pei, Jen-Wei Huang, Jerry Chun-Wei Lin.

Format:

Book

Contributor:

Yang, De-Nian, editor.

Series:

Lecture Notes in Artificial Intelligence, 2945-9141 ; 14649

Language:

English

Subjects (All):

Artificial intelligence.

Algorithms.

Education--Data processing.

Education.

Computer science--Mathematics.

Computer science.

Signal processing.

Computer networks.

Artificial Intelligence.

Design and Analysis of Algorithms.

Computers and Education.

Mathematics of Computing.

Signal, Speech and Image Processing.

Computer Communication Networks.

Local Subjects:

Artificial Intelligence.

Design and Analysis of Algorithms.

Computers and Education.

Mathematics of Computing.

Signal, Speech and Image Processing.

Computer Communication Networks.

Physical Description:

1 online resource (431 pages)

Edition:

1st ed. 2024.

Place of Publication:

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

Summary:

The 6-volume set LNAI 14645-14650 constitutes the proceedings of the 28th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2024, which took place in Taipei, Taiwan, during May 7–10, 2024. The 177 papers presented in these proceedings were carefully reviewed and selected from 720 submissions. They deal with new ideas, original research results, and practical development experiences from all KDD related areas, including data mining, data warehousing, machine learning, artificial intelligence, databases, statistics, knowledge engineering, big data technologies, and foundations.

Contents:

Intro

General Chairs' Preface

PC Chairs' Preface

Organization

Contents - Part V

Multimedia and Multimodal Data

Re-thinking Human Activity Recognition with Hierarchy-Aware Label Relationship Modeling

1 Introduction

2 Related Work

2.1 Human Activity Recognition (HAR)

2.2 Hierarchical Label Modeling

3 Problem Formulation

4 Our Proposals

4.1 Hierarchy-Aware Label Encoding

4.2 Activity Data Encoding

4.3 Label-Data Joint Embedding Learning

5 Experiments

5.1 Experimental Settings

5.2 Experimental Results

5.3 Ablation Study

6 Discussions and Conclusion

References

Geometrically-Aware Dual Transformer Encoding Visual and Textual Features for Image Captioning

2 Related Works

3 Proposed Approach

3.1 Features Extractor

3.2 Caption Generator

3.3 Attention Block

3.4 Training and Objectives

4 Experiments

4.1 Experiments Setup

4.2 Experiment Result

5 Conclusions

MHDF: Multi-source Heterogeneous Data Progressive Fusion for Fake News Detection

3 MHDF Model

3.1 Model Overview

3.2 Multi-source Heterogeneous Data Amplification

3.3 News Textual Feature Fusion

3.4 News Visual Feature Fusion

3.5 Sentiment Feature Extractor

3.6 Feature Integration Classifier

4.1 Dataset

4.2 Experimental Settings

4.3 Performance Comparison

4.4 Ablation Experiments and Validity Verification

4.5 Conclusions

Accurate Semi-supervised Automatic Speech Recognition via Multi-hypotheses-Based Curriculum Learning

2.1 Automatic Speech Recognition Methods

2.2 Connectionist Temporal Classification (CTC) Loss

3 Proposed Method

3.1 Multiple Hypotheses for Unlabeled Instances.

3.2 Training ASR Model with Multiple Hypotheses

3.3 Curriculum Learning

3.4 Theoretical Analysis

4.1 Experimental Settings

4.2 Transcription Performance (Q1)

4.3 Speed of Convergence (Q2)

4.4 Ablation Study (Q3)

MM-PhyQA: Multimodal Physics Question-Answering with Multi-image CoT Prompting

2.1 Available Datasets

2.2 Large Multimodal Models and Chain-of-Thought

3 Novel Dataset

3.1 Original Dataset Creation

3.2 Data Augmentation Procedure

3.3 Chain of Thought Variant

3.4 MM-PhyQA Dataset Topics

4 Methodology

4.1 Multi-image Chain-of-Thought (MI-CoT)

5.1 Models

6 Results and Discussion

6.1 Model Performance

6.2 Zero Shot Prompting Vs Supervised Fine-Tuning

6.3 Effect of Chain of Thought Prompting

6.4 Error Analysis

7 Conclusion

Adversarial Text Purification: A Large Language Model Approach for Defense

3 Background

3.1 Large Language Models

3.2 Adversarial Text Purification

4 LLM-Guided Adversarial Text Purification

5.1 Experimental Setting

5.2 Results and Discussion

6 Conclusion

lil'HDoC: An Algorithm for Good Arm Identification Under Small Threshold Gap

2 Background

2.1 Good Arm Identification

3 Problem Setting

4 Preliminary

5 Algorithm

5.1 Correctness of lil'HDoC

5.2 First Arms Sampling Complexity

5.3 Total Sample Complexity

6 Experiment

6.1 Dataset

6.2 Baseline

6.3 Results

Recommender Systems

ScaleViz: Scaling Visualization Recommendation Models on Large Data

4 Proposed Solution

4.1 Cost Profiling

4.2 RL Agent.

5 Evaluations

5.1 Experimental Setup

5.2 Speed-Up in Visualization Generation

5.3 Budget vs. Error Trade-Off

5.4 Need for Dataset-Specific Feature Selection

5.5 Scalability with Increasing Data Size

Collaborative Filtering in Latent Space: A Bayesian Approach for Cold-Start Music Recommendation

2 Related Work and Problem Formulation

2.1 Problem Formulation

3 Methodology

3.1 Overview

3.2 Statistical Model in CFLS

3.3 Optimization

3.4 Prediction

4.3 Performance Comparisons

4.4 Influence of Different Cold-Start Levels

4.5 Diversity, Interpretability and User Controllability

On Diverse and Precise Recommendations for Small and Medium-Sized Enterprises

3 Definitions and Problem Statement

4 Variants of a Session-Based Recommender System

4.1 Quality Metrics

5 Experiments and Evaluation

5.1 Selection of Real-World Datasets

5.2 Task Definition and Parameter Configuration

5.3 Evaluation of Experimental Results

6 Conclusion and Future Work

HMAR: Hierarchical Masked Attention for Multi-behaviour Recommendation

2 Methodology

2.2 HMAR

2.3 Multi-task Learning

3 Experiments

3.1 Experimental Settings

3.2 Evaluation Protocol

3.3 Model Performance (RQ1)

3.4 Effect of Auxiliary Behaviors and Individual Model Components (RQ2 &amp

RQ3)

4 Related Work

5 Conclusion

Residual Spatio-Temporal Collaborative Networks for Next POI Recommendation

3 Method

3.1 Problem Formulation

3.2 Long-Term Dependence Module

3.3 Short-Term Dependence Module

3.4 Sample Balancer.

4 Experiments

4.2 Recommendation Performance

4.3 Ablation Study

Conditional Denoising Diffusion for Sequential Recommendation

3.1 Stepwise Diffuser

3.2 Sequence Encoder

3.3 Cross-Attentive Conditional Denoising Decoder

3.4 Optimization

4.1 Plateau of Ranking Prediction

4.2 Overall Experiments

4.4 Hyperparameter Sensitivity

4.5 Case Study for Stepwise Generation

UIPC-MF: User-Item Prototype Connection Matrix Factorization for Explainable Collaborative Filtering

2.1 Collaborative Filtering

2.2 Explainable and Transparent Recommender Models

2.3 The Prototype-Based Collaborative Filtering

3.1 User-Item Prototypes Connections Matrix Factorization (UIPC-MF)

3.2 Loss Function

4 Experiments and Discussion

4.1 Evaluation Metrics

4.2 Baseline Models

4.3 Training Details

4.4 Evaluation Results

4.5 Explaining UIPC-MF Recommendations

4.6 The Impact of L1-Norm in Reduction of Learning Bias

Towards Multi-subsession Conversational Recommendation

3 MSMCR Scenario

3.1 Definition

3.2 General Framework

4.1 Context-Aware Recommendation

4.2 Policy Learning

4.3 Model Training

5.2 Overall Performance

5.3 Further Experiments

False Negative Sample Aware Negative Sampling for Recommendation

3 Preliminary

4.1 False Negatives Identification

4.2 False Negatives Elimination

5 Experiment

5.1 Experiment Settings.

5.2 Performance Comparison

5.3 Study of EDNS

Multi-sourced Integrated Ranking with Exposure Fairness

2 Problem Formulation

3 Proposed Model

3.1 Input Layer

3.2 Dual RNN Module

3.3 Multi-task Module

3.4 Model Training

4.2 Baselines

4.3 Model Selection

4.4 Performance Comparison

4.5 Ablation Study

4.6 Online A/B Testing

Soft Contrastive Learning for Implicit Feedback Recommendations

3.1 Notations

3.2 The SCLRec Framework

4.2 Overall Performance (RQ1)

4.3 Ablation Study (RQ2)

4.4 Robustness to Interaction Noises (RQ3)

Dual-Graph Convolutional Network and Dual-View Fusion for Group Recommendation

3 Approach

3.1 Dual-Graph Construction

3.2 Dual-Graph Network for Member Preference

3.3 Dual-View Fusion for Group Preference

3.4 Group Recommendation and Model Training

4.1 Experimental Dataset and Setup

4.2 Experimental Results and Analysis

4.3 Parameter Sensitivity

5 Related Works

TripleS: A Subsidy-Supported Storage for Electricity with Self-financing Management System

2 Literature Review

2.1 Electricity Subsidy and Operating Reserve

2.2 Electricity Management System

2.3 Electricity Storage

3 Problem Definition and Simulation Environment

4 Proposed TripleS

5 Experimental Results

5.1 Performance Evaluation

5.2 Performance Evaluation Under MS Attack

5.3 Influence of Self-discharge

Spatio-temporal Data.

Mask Adaptive Spatial-Temporal Recurrent Neural Network for Traffic Forecasting.

Notes:

Includes bibliographical references and index.

ISBN:

9789819722624

9819722624