Advances in Knowledge Discovery and Data Mining

Name: Advances in Knowledge Discovery and Data Mining | 27th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2023, Osaka, Japan, May 25-28, 2023, Proceedings, Part II
Brand: Springer
Price: 139.09 EUR
Availability: OnlineOnly

27th Pacific-Asia Conference on Knowledge Discovery and Data Mining, PAKDD 2023, Osaka, Japan, May 25-28, 2023, Proceedings, Part II

Hisashi Kashima Tsuyoshi Ide Wen-Chih Peng(Herausgeber*in)

Springer (Verlag)

Erschienen am 27. Mai 2023

XVII, 567 Seiten

E-Book

PDF mit Wasserzeichen-DRM

Systemvoraussetzungen

978-3-031-33377-4 (ISBN)

139,09 €inkl. 7% MwSt.

Systemvoraussetzungen

für PDF mit Wasserzeichen-DRM

E-Book Einzellizenz

Als Download verfügbar

Beschreibung

Weitere Details

Weitere Ausgaben

Inhalt

Intro
General Chairs' Preface
PC Chairs' Preface
Organization
Contents - Part II
Graphs and Networks
Improving Knowledge Graph Entity Alignment with Graph Augmentation
1 Introduction
2 Related Works
3 Preliminaries
4 Methodology
4.1 Entity-Relation Encoder
4.2 Model Training with Graph Augmentation
4.3 Alignment Inference
5 Experimental Setup
5.1 Experimental Setup
5.2 Experimental Results
6 Discussion and Conclusion
References
MixER: MLP-Mixer Knowledge Graph Embedding for Capturing Rich Entity-Relation Interactions in Link Prediction
1 Introduction
2 Related Work
2.1 Translation-Based Approaches
2.2 Matrix Factorization-Based Approaches
2.3 Neural Network-Based Approaches
3 Methodology
3.1 Problem Formulation and Notations
3.2 Overall Architecture Design
3.3 Model Architecture
4 Experiments
4.1 Datasets
4.2 Evaluation Protocol and Metric
4.3 Hyperparameters and Baselines
4.4 Results and Discussion
4.5 Analysis
5 Conclusion and Future Work
References
GTEA: Inductive Representation Learning on Temporal Interaction Graphs via Temporal Edge Aggregation
1 Introduction
2 Related Works
2.1 Temporal Dynamics Modeling on Graph-Structured Data
2.2 Representation Learning on Graphs with Edge Features
3 Proposed Methods
3.1 Problem Formulation
3.2 Overview of GTEA
3.3 Learning Edge Embeddings for Interaction Sequences
3.4 Representation Learning with Temporal Edge Aggregation
3.5 Model Training for Different Graph-Related Tasks
4 Experiments
4.1 Experimental Setup
4.2 Experimental Results of Overall Performance
4.3 Experiments Analyses
5 Conclusions
References
You Need to Look Globally: Discovering Representative Topology Structures to Enhance Graph Neural Network
1 Introduction
2 Related Works
3 Problem Formulation
4 Methodology
4.1 Global Topology Structure Extraction
4.2 Graph Structure Memory Augmented Representation Learning
4.3 Objective Function of GSM-GNN
5 Experiments
5.1 Datasets
5.2 Experimental Setup
5.3 Performance on Node Classification
5.4 Flexibility of GSM-GNN for Various GNNs
5.5 Ablation Study
6 Conclusion
References
UPGAT: Uncertainty-Aware Pseudo-neighbor Augmented Knowledge Graph Attention Network
1 Introduction
2 Preliminaries
2.1 Problem Statement
2.2 Motivations and Challenges
2.3 Related Work
3 Approach
3.1 Overview
3.2 1-Hop Attention Module with Attention Baseline Mechanism
3.3 Confidence Score Prediction and Training Objective
3.4 Pseudo-neighbor Augmented Graph Attention Network
4 Experiment
4.1 Settings
4.2 Results and Analysis
4.3 Ablation Study
4.4 Deterministic Settings
5 Conclusion and Future Work
References
Mining Frequent Sequential Subgraph Evolutions in Dynamic Attributed Graphs
1 Introduction
2 Related Work
3 Notations
3.1 Dynamic Attributed Graph
3.2 A New Pattern Domain
3.3 Interesting Measures and Constraints
4 Mining Frequent Sequential Subgraph Evolutions
4.1 Extraction of Subgraph Candidates
4.2 Generation of Size-1 Patterns by Graph Addition
4.3 Extension of Patterns
5 Experiments
6 Conclusion
References
CondTraj-GAN: Conditional Sequential GAN for Generating Synthetic Vehicle Trajectories
1 Introduction
2 Problem Definition
3 The CondTraj-GAN Framework
3.1 Training
3.2 Trajectory Inference
4 Evaluation Setup
4.1 Dataset
4.2 Model Setups
4.3 Evaluation Metrics
4.4 Baselines
5 Evaluation
5.1 Trajectory Generation Performance
5.2 Ablation Study
6 Related Work
7 Conclusion and Future Work
References
A Graph Contrastive Learning Framework with Adaptive Augmentation and Encoding for Unaligned Views
1 Introduction
2 Related Work
2.1 Graph Contrastive Learning
2.2 Adversarial Training
3 Method
3.1 Preliminaries
3.2 Adaptive Augmentation
3.3 Encoding Methods for Homophilic and Heterophilic Graphs
3.4 G-EMD-based Contrastive Loss
3.5 Adversarial Training on GCAUV
4 Experiments
4.1 Experimental Setup
4.2 Performance on Node Classification
4.3 Ablation Studies
5 Conclusion
References
MPool: Motif-Based Graph Pooling
1 Introduction
2 Related Work
3 Methodology
3.1 Preliminaries and Problem Formulation
3.2 Motif Based Graph Pooling Models
3.3 Readout Function and Output Layer
4 Experiment
4.1 Overall Evaluation
5 Conclusion
References
Anti-Money Laundering in Cryptocurrency via Multi-Relational Graph Neural Network
1 Introduction
2 Methodology
2.1 Graph Construction
2.2 Representation Embedding
2.3 Inter-relation Aggregation
2.4 Adaptive Neighbor Sampler
3 Experiment
3.1 Experimental Setup
3.2 Demystifying Mixing Behavior
3.3 Performance Comparison
3.4 Ablation Study
3.5 Adaptive Sampler Analysis
4 Conclusion
References
Interpretability and Explainability
CeFlow: A Robust and Efficient Counterfactual Explanation Framework for Tabular Data Using Normalizing Flows
1 Introduction
2 Related Works
3 Preliminaries
3.1 Counterfactual Explanation
3.2 Normalizing Flow
4 Methodology
4.1 General Architecture of CeFlow
4.2 Normalizing Flows for Categorical Features
4.3 Conditional Flow Gaussian Mixture Model for Tabular Data
4.4 Counterfactual Generation Step
5 Experiments
6 Conclusion
References
Feedback Effect in User Interaction with Intelligent Assistants: Delayed Engagement, Adaption and Drop-out
1 Introduction
2 Related Work
3 Data Collection
3.1 Study 1: Pre-event Control Period
3.2 Study 2: Post-event New User Period
4 Feedback Effect on Engagement
4.1 Covariates and Outcome Variables
4.2 Observational Causal Methods
4.3 Time to Next Engagement
4.4 Number of Active Days
5 Language Convergence in New User Cohort
5.1 New and Existing User Cohort Definition
5.2 New User's Self-Selection: Drop-out or Adaption
6 Discussions
References
Toward Interpretable Machine Learning: Constructing Polynomial Models Based on Feature Interaction Trees
1 Introduction
2 Related Work
2.1 SHAP and Pair-Wise Interaction Values
2.2 Polynomial Model and EBM
3 Methodology
3.1 Black-box Model Creation
3.2 Global SHAP Interaction Value Score Calculation
3.3 Tree-building Process
4 Experiments
4.1 Model Performance
4.2 Evaluating Interpretability
4.3 Usability Study
5 Conclusion
References
Kernel Methods
BioSequence2Vec: Efficient Embedding Generation for Biological Sequences
1 Introduction
2 Related Work
3 Proposed Approach
3.1 BioSequence2Vec Representation
4 Experimental Evaluation
5 Results and Discussion
6 Conclusion
References
Matrices and Tensors
Relations Between Adjacency and Modularity Graph Partitioning
1 Introduction
2 Preliminaries
3 Dominant Eigenvectors of Modularity and Adjacency Matrices
4 Normalized Adjacency and Modularity Clustering
5 Experiments
5.1 Synthetic Data Sets
5.2 PenDigit Data Sets from MNIST Database
6 Conclusion
References
Model Selection and Evaluation
Bayesian Optimization over Mixed Type Inputs with Encoding Methods
1 Introduction
2 Related Work
2.1 BO for Categorical and Continuous Inputs
2.2 Encoding Methods
3 Background
3.1 Problem Statement
3.2 Bayesian Optimization
4 The Proposed Framework
4.1 Target Mean Encoding BO
4.2 Aggregate Ordinal Encoding BO
5 Experiments
5.1 Baseline Method and Evaluation Measures
5.2 Performance and Computation Time
6 Conclusion
References
Online and Streaming Algorithms
Using Flexible Memories to Reduce Catastrophic Forgetting
1 Introduction
2 Related Work
3 The Continual Learning Problem
4 The Stability Wrapper (SW) for Replay Buffer Replacements
5 Experimental Results
6 Conclusion
References
Fair Healthcare Rationing to Maximize Dynamic Utilities
1 Introduction
1.1 Our Models
1.2 Our Contributions
2 Algorithms for Model 1
2.1 Online Algorithm for Model 1
2.2 Charging Scheme
2.3 Tight Example for the Online Algorithm
3 Online Algorithm for Model 2
3.1 Outline of the Charging Scheme
4 Strategy-Proofness of the Online Algorithm
5 Experimental Evaluation
5.1 Methodology
5.2 Datasets
5.3 Results and Discussions
6 Conclusion
References
A Multi-player MAB Approach for Distributed Selection Problems
1 Introduction
2 Related Work
3 Platform Model and Problem Formulation
4 The Offline Optimization Problem
5 Online Learning Algorithm
6 Experiment
7 Conclusion
References
A Thompson Sampling Approach to Unifying Causal Inference and Bandit Learning
1 Introduction
2 Model
2.1 The Bandit Learning Model
2.2 The Data Model
2.3 Problem Formulation
3 Limitations of Naively Applying Thompson Sampling
3.1 VirTS: Naively Applying Thompson Sampling
3.2 Limitations of VirTS
4 VirTS-DF: Improving VirTS via Offline Data Filtering
5 Experiments on Real-world Data
5.1 Experimental Settings
5.2 Experiment Results
6 Related Work
7 Conclusion
References
Parallel and Distributed Mining
Maverick Matters: Client Contribution and Selection in Federated Learning
1 Introduction
2 Related Studies
3 Federated Learning with Mavericks
3.1 Shapley Value for Mavericks
4 FedEMD
5 Experimental Evaluation
5.1 FedEMD Is Effective for Client Selection
5.2 FedEMD Works for Multiple Mavericks
6 Conclusion
References
pFedV: Mitigating Feature Distribution Skewness via Personalized Federated Learning with Variational Distribution Constraints*-12pt
1 Introduction
2 Related Work and Background
2.1 Federated Learning
2.2 Statistical Heterogeneity
2.3 Variational Inference
3 Methodology
3.1 Problem Formulation
3.2 Derivation of Variational Distribution Constraints
3.3 Personalized Federated Learning with Variational Distribution Constraints
4 Experiments
4.1 Experimental Settings
4.2 Results
5 Conclusion
References
Probabilistic Models and Statistical Inference
Inverse Problem of Censored Markov Chain: Estimating Markov Chain Parameters from Censored Transition Data*-12pt
1 Introduction
2 Related Works
3 Markov Chain and Censored Markov Chain
4 Estimating Markov Chain Parameters from Censored Transition Data
4.1 Problem Formulation
4.2 Difficulty of Our Problem
4.3 Parameter Estimation via Divergence Minimization
5 Experiment
5.1 Setting
5.2 Results
6 Conclusion
References
Parameter-Free Bayesian Decision Trees for Uplift Modeling
1 Introduction
2 Context and Literature Overview
2.1 Uplift Problem Formulation
2.2 Related Work
3 UB-DT: Uplift Decision Tree Approach
3.1 Parameters of an Uplift Tree Model T
3.2 Uplift Tree Evaluation Criterion
3.3 C(T): Proof of Equation2
3.4 Search Algorithm
3.5 UB-RF
4 Experiments
4.1 Is UB-DT a Good Uplift Estimator?
4.2 UB-DT and UB-RF Versus State of the Art Methods
5 Conclusion and Perspectives
References
Interpretability Meets Generalizability: A Hybrid Machine Learning System to Identify Nonlinear Granger Causality in Global Stock Indices
1 Introduction
2 Linear Granger Causality
3 Related Work
3.1 Machine Learning Methods
3.2 Granger Causality in Finance
4 Methodology
4.1 Module 1: Nonlinear Granger Causality
4.2 Module 2: Time Series Prediction
4.3 Module 3: Significance and Negotiation
5 Experiments and Results
5.1 Validating the Methods
5.2 Application of the Stock Indices
6 Conclusion
References
Reinforcement Learning
A Dynamic and Task-Independent Reward Shaping Approach for Discrete Partially Observable Markov Decision Processes
1 Introduction
2 Background and Notation
3 Related Work
4 LDPG: Our Proposed Reward Shaping Method
5 Experimental Results
5.1 Results
6 Conclusion and Future Work
References
Multi-Agent Meta-Reinforcement Learning with Coordination and Reward Shaping for Traffic Signal Control
1 Introduction
2 Problem Definition
3 Method
3.1 Coordination Strategy
3.2 Reward Design
3.3 Procedure of Meta-CSLight
4 Experiments
4.1 Datasets
4.2 Baseline Methods
4.3 Comparison with Baseline Methods
5 Conclusion
References
Regularization of the Policy Updates for Stabilizing Mean Field Games
1 Introduction
2 Related Works
3 Problem Formulation
4 Proposal: Proximal Policy Updates for MFG
5 Experimentation
5.1 Experimental Setup
5.2 Numerical Results
5.3 Analysis on the Hyper-parameters
6 Conclusion
References
Constrained Portfolio Management Using Action Space Decomposition for Reinforcement Learning
1 Introduction
2 Related Work
3 Problem Setting
4 Solution as CMDP
5 Action Space Decomposition Based Optimization
6 Experiments
7 Conclusion
References
Transfer Reinforcement Learning Based Negotiating Agent Framework
1 Introduction
2 Preliminaries
2.1 Negotiation Settings
2.2 Reinforcement Learning
3 Transfer Learning Based Agent
3.1 Framework Overview
3.2 Negotiation Module
3.3 Adaptation Module
3.4 Transfer Module
4 Experiments
4.1 New Opponent Learning Task
4.2 Performance Against ANAC Winning Agents
5 Conclusion and Future Work
References
Relational Learning
A Relational Instance-Based Clustering Method with Contrastive Learning for Open Relation Extraction
1 Introduction
2 Related Work
3 Methodology
3.1 Relational Instance Encoder
3.2 Instance-Relational Contrastive Learning
3.3 Clustering
3.4 Relation Classification
4 Experimental Setup
4.1 Datasets
4.2 Baseline and Evaluation Metrics
4.3 Implementation Details
5 Results and Analysis
5.1 Main Results
5.2 Ablation Study
5.3 Visualization and Analysis
6 Conclusions
References
Security and Privacy
Achieving Provable Byzantine Fault-tolerance in a Semi-honest Federated Learning Setting
1 Introduction
2 Related Work
3 The Proposed Method
3.1 The Setting and Threat Model
3.2 FedPBF
4 Experiments
4.1 Setup and Evaluation Metrics
4.2 Comparison with Existing Byzantine Fault-tolerant Methods
4.3 Robustness
5 Conclusion
References
Defending Against Backdoor Attacks by Layer-wise Feature Analysis
1 Introduction
2 Related Work: Backdoor Attacks and Defenses
3 Layer-wise Feature Analysis
4 The Proposed Defense
5 Experiments
5.1 Main Settings
5.2 Main Results
5.3 Discussions
6 Conclusion
References
Enhancing Federated Learning Robustness Using Data-Agnostic Model Pruning
1 Introduction
2 Related Work
3 Problem Statement
3.1 Federated Learning
3.2 Threat Model
4 Proposed FLAP
4.1 Approach Overview
4.2 Model Pruning
5 Evaluation
5.1 Experiment Setup
5.2 Benign FL with FLAP
5.3 FLAP in Adversarial Settings
6 Discussion
7 Conclusion
References
BeamAttack: Generating High-quality Textual Adversarial Examples Through Beam Search and Mixed Semantic Spaces
1 Introduction
2 Related Work
3 Beam Search Adversarial Attack
3.1 Black-box Untargeted Attack
3.2 Word Importance Calculation
3.3 Mixed Semantic Spaces
3.4 Improved Beam Search
4 Experiments
4.1 Experimental Results
4.2 Ablation Study
4.3 Transferability
4.4 Adversarial Training
5 Conclusion
References
Semi-supervised and Unsupervised Learning
Reduction from Complementary-Label Learning to Probability Estimates
1 Introduction
2 Problem Setup
2.1 Ordinary-label Learning
2.2 Complementary-label Learning
3 Proposed Framework
3.1 Motivation
3.2 Methodology
3.3 Connection to Previous Methods
4 Experiments
4.1 Experiment Setup
4.2 Discussion
4.3 Learn from CL with Traditional Methods
5 Conclusion
References
Semi-Supervised Text Classification via Self-Paced Semantic-Level Contrast*-12pt
1 Introduction
2 Related Work
2.1 Semi-Supervised Text Classification
2.2 Contrastive Learning
3 Methodology
3.1 Overview of S2CL
3.2 Self-Paced Pseudo-Label Generator
3.3 Robust Supervised Learning
3.4 Semantic-Level Contrastive Learning
3.5 Connecting RSL and SLCL
4 Experiments
4.1 Datasets and Baselines
4.2 Benchmarking S2CL on Public Datasets
4.3 Benchmarking S2CL on Events39
4.4 Deepgoing Exploration
5 Conclusion
References
Multi-Augmentation Contrastive Learning as Multi-Objective Optimization for Graph Neural Networks*-12pt
1 Introduction
2 Related Work
2.1 Graph Contrastive Learning
2.2 Multi-objective Optimization
3 Methodology
3.1 Model Overview
3.2 Specific Augmentation Contrastive Loss
3.3 Instantiation of PGCL as Multi-Objective Optimization
3.4 Augmentation-Dependent Embedding Subspace
4 Experiments
4.1 Conflict Among Augmentations
4.2 Unsupervised Representation Learning
5 Conclusion
References
Adversarial Active Learning with Guided BERT Feature Encoding*-12pt
1 Introduction
2 Related Work
3 Learning Framework
3.1 Problem Formulation
3.2 Proposed Framework: B-ASAL
3.3 Learning Objective
4 Experiments
4.1 Question Answering Classification
4.2 Multi-Genre Natural Language Inference
4.3 Multi-Label Emotion Classification
4.4 Further Analysis
5 Conclusions and Future Work
References
Theoretical Foundations
Accelerating Stochastic Newton Method via Chebyshev Polynomial Approximation
1 Introduction
2 Background
2.1 Estimator for the Hessian Inverse in LiSSA
2.2 Chebyshev Polynomial Approximation
3 Our Proposed CP-SNM
4 Theoretical Results
5 Experiments
6 Conclusion
References
Stochastic Submodular Maximization via Polynomial Estimators
1 Introduction
2 Related Work
3 Technical Preliminary
3.1 Problem Definition
3.2 Change of Variables and Multiliear Relaxation
3.3 Stochastic Continuous Greedy Algorithm
3.4 Multilinear Functions and the Multilinear Relaxation of a Polynomial
4 Main Results
4.1 Polynomial Estimator
5 Problem Examples
5.1 Data Summarization (SM) ch41lin2011class,ch41mirzasoleiman2016fast,ch41kazemi2019submodular
5.2 Influence Maximization (IM) ch41kempe2003maximizing,ch41chen2009efficient
5.3 Facility Location (FL) ch41mokhtari2018conditional
6 Experiments
7 Conclusions
References
Transfer Learning and Meta Learning
Few-Shot Human Motion Prediction for Heterogeneous Sensors
1 Introduction
2 Related Work
3 Methodology
3.1 Problem Definition
3.2 GraphHetNet
4 Results
4.1 Experimental Setup
4.2 Results
4.3 Ablations
5 Conclusion
References
Correction to: Anti-Money Laundering in Cryptocurrency via Multi-Relational Graph Neural Network
Correction to: Chapter "Anti-Money Laundering in Cryptocurrency via Multi-Relational Graph Neural Network" in: H. Kashima et al. (Eds.): Advances in Knowledge Discovery and Data Mining, LNAI 13936, https://doi.org/10.1007/978-3-031-33377-4_10
Author Index

Systemvoraussetzungen

Als PDF speichern Als Link merken