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 VI / edited by De-Nian Yang, Xing Xie, Vincent S. Tseng, Jian Pei, Jen-Wei Huang, Jerry Chun-Wei Lin.

Format:

Book

Author/Creator:

Yang, De-Nian.

Contributor:

Xie, Xing.

Tseng, Vincent S.

Pei, Jian.

Huang, Renwei.

Lin, Jerry Chun-Wei.

Series:

Lecture Notes in Artificial Intelligence, 2945-9141 ; 14650

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 (329 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 VI

Scientific Data

FR3LS: A Forecasting Model with Robust and Reduced Redundancy Latent Series

1 Introduction

2 Related Work

3 Problem Setup

4 Model Architecture

4.1 Temporal Contextual Consistency

4.2 Non-contrastive Representations Learning

4.3 Deterministic Forecasting

4.4 Probabilistic Forecasting

4.5 End-to-End Training

5 Experiments

5.1 Experimental Results

5.2 Visualization of Latent and Original Series Forecasts

5.3 Further Experimental Setup Details

6 Conclusion

References

Knowledge-Infused Optimization for Parameter Selection in Numerical Simulations

2 Preliminaries on Numerical Simulations

3 Identification and Analysis of Relevant Metadata

4 Efficient Metadata Capture for Parameter Optimization

4.1 Early Termination of Farming Runs

4.2 PROBE: Probing Specific Parameter Combinations

5 Experimental Evaluation

5.1 Quality of Parameter Optimization Using PROBE

5.2 Efficiency Evaluation

5.3 Reuse of Metadata Acquired Through PROBE

5.4 Generalization to Other Model Problems and Schemes

6 Conclusion and Outlook

Material Microstructure Design Using VAE-Regression with a Multimodal Prior

2 Methodology

3 Related Work

4 Experimental Results

5 Summary and Conclusions

A Weighted Cross-Modal Feature Aggregation Network for Rumor Detection

2.1 Rumor Detection

2.2 Multimodal Alignment

3 Methodology

3.1 Overview of WCAN

3.2 Feature Extraction

3.3 Weighted Cross-Modal Aggregation Module

3.4 Multimodal Feature Fusion

3.5 Objective Function

4 Experiments

4.1 Datasets

4.2 Baselines

4.3 Ablation Experiment.

4.4 Hyper-parameter Analysis

4.5 Visualization on the Representations

4.6 Case Study

5 Conclusions

Texts, Web, Social Network

Quantifying Opinion Rejection: A Method to Detect Social Media Echo Chambers

3 Preliminaries

4 Echo Chamber Detection in Signed Networks

5 SEcho Method

5.1 SEcho Metric

5.2 Greedy Optimisation

6 Experiments

7 Conclusion

KiProL: A Knowledge-Injected Prompt Learning Framework for Language Generation

2.1 Problem Statement

2.2 Knowledge-Injected Prompt Learning Generation

2.3 Training and Inference

3 Experiments

3.1 Datasets

3.2 Settings

3.3 Automatic Evaluation

3.4 Human Annotation

3.5 Ablation Study

3.6 In-Depth Analysis

3.7 Case Study

4 Conclusion

GViG: Generative Visual Grounding Using Prompt-Based Language Modeling for Visual Question Answering

2.1 Pix2Seq Framework

2.2 Prompt Tuning

3.1 Prompt Tuning Module

3.2 VG Module

3.3 Conditional Trie-Based Search Algorithm (CTS)

4 Results

4.1 Dataset Description

4.2 Results on WSDM 2023 Toloka VQA Dataset Benchmark

5 Discussion

5.1 Prompt Study

5.2 Interpretable Attention

Aspect-Based Fake News Detection

3.1 Problem Definition and Model Overview

3.2 Aspect Learning and Extraction

3.3 News Article Classification

4.2 Experimental Settings

4.3 Evaluation

5 Analysing the Effect of Aspects Across Topics

6 Discussion and Future Work

DQAC: Detoxifying Query Auto-completion with Adapters

3 Methodology.

3.1 QDetoxify: Toxicity Classifier for Search Queries

3.2 The DQAC Model

4 Experimental Setup

5 Results and Analyses

6 Conclusions

Graph Neural Network Approach to Semantic Type Detection in Tables

2 Related Works

3 Problem Definition

4 GAIT

4.1 Single-Column Prediction

4.2 Graph-Based Prediction

4.3 Overall Prediction

5 Evaluation

5.1 Evaluation Method

5.2 Results

TCGNN: Text-Clustering Graph Neural Networks for Fake News Detection on Social Media

3 The Proposed TCGNN Method

3.1 Text-Clustering Graph Construction

3.2 Model Training

5 Conclusions and Discussion

Exploiting Adaptive Contextual Masking for Aspect-Based Sentiment Analysis

3.1 Problem Formulation and Motivation

3.2 Standalone ABSA Tasks

3.3 Adaptive Contextual Threshold Masking (ACTM)

3.4 Adaptive Attention Masking (AAM)

3.5 Adaptive Mask Over Masking (AMOM)

3.6 Training Procedure for ATE and ASC

4 Experiments and Results

5 Conclusion

An Automated Approach for Generating Conceptual Riddles

3.1 Triples Creator

3.2 Properties Classifier

3.3 Generator

3.4 Validator

4 Evaluation and Results

5 Conclusion and Future Work

Time-Series and Streaming Data

DiffFind: Discovering Differential Equations from Time Series

2 Background and Related Work

2.1 Related Work

2.2 Background - Genetic Algorithms for Architecture Search

3 Proposed Method: DiffFind

4.1 Q1 - DiffFind is Effective

4.2 Q2 - DiffFind is Explainable

4.3 Q3 - DiffFind is Scalable

References.

DEAL: Data-Efficient Active Learning for Regression Under Drift

3 Problem Statement and Notation

4 Our Method: DEAL

4.1 The Adapted Stream-Based AL Cycle

4.2 Our Drift-Aware Estimation Model

5 Experimental Design

5.1 Baselines

5.2 Evaluation Data

5.3 Evaluation Metrics

6 Evaluation

6.1 Comparison of DEAL Against Baselines

6.2 Impact of the User-Required Error Threshold

Evolving Super Graph Neural Networks for Large-Scale Time-Series Forecasting

2 Related Models

3 Evolving Super Graph Neural Networks

3.1 Preliminary Notations

3.2 Super Graph Construction

4 Diffusion on Evolving Super Graphs

4.1 Predictor

5 Experiments on Large-Scale Datasets

5.1 Forecasting Result and Analysis

5.2 Runtime and Space Usage Analysis

5.3 Ablation Study

Unlearnable Examples for Time Series

2.1 Data Poisoning

2.2 Adversarial Attack

2.3 Unlearnable Examples

3 Error-Minimizing Noise for Time Series

3.1 Objective

3.2 Threat Model

3.3 Challenges

3.4 Problem Formulation

3.5 A Straightforward Baseline Approach

3.6 Controllable Noise on Partial Time Series Samples

4.1 Experiment Setup

4.2 Against Classification Models

4.3 Against Generative Models

Learning Disentangled Task-Related Representation for Time Series

3 The Proposed Method

3.1 Overview

3.2 Task-Relevant Feature Disentangled

3.3 Task-Adaptive Augmentation Selection

4 Experiments and Discussions

4.1 Datasets and Implementation Details

4.2 Ablation Analysis

4.3 Results on Classification Tasks

4.4 Results on Forecasting Tasks

4.5 Visualization Analysis.

5 Conclusion

A Multi-view Feature Construction and Multi-Encoder-Decoder Transformer Architecture for Time Series Classification

3 Problem Formulation

4 Methodology

4.1 Feature Construction

4.2 Multi-view Representation

4.3 Multi-Encoder-Decoder Transformer (MEDT) Classification

5.1 Experiments Using Multivariate Time Series Data Benchmarks

5.2 Experiment Using a Real-World Physical Activities Dataset

Kernel Representation Learning with Dynamic Regime Discovery for Time Series Forecasting

3.1 Key Concepts

3.2 Self-representation Learning in Time Series

3.3 Kernel Trick for Modeling Time Series

4 Proposed Method

4.1 Kernel Representation Learning: Modeling Regime Behavior

4.2 Forecasting

5.1 Data

5.2 Experimental Setup and Evaluation

5.3 Regime Identification

5.4 Benchmark Comparison

5.5 Ablation Study

Hyperparameter Tuning MLP's for Probabilistic Time Series Forecasting

2 Problem Statement

3 MLPs for Time Series Forecasting

3.1 Nlinear Model

4 Hyperparameters

4.1 Time Series Specific Configuration

4.2 Training Specific Configurations

4.3 TSBench-Metadataset

5 Experimental Setup

6 Results

Efficient and Accurate Similarity-Aware Graph Neural Network for Semi-supervised Time Series Classification

2.1 Graph-Based Time Series Classification

2.2 Lower Bound of DTW

4.1 Batch Sampling

4.2 LB_Keogh Graph Construction

4.3 Graph Convolution and Classification

4.4 Advantages of Our Model

5 Experimental Evaluation.

5.1 Comparing with 1NN-DTW.

Notes:

Description based on publisher supplied metadata and other sources.

ISBN:

9789819722662

9819722667

OCLC:

1432236147