Machine Learning and Knowledge Discovery in Databases. Research Track

Name: Machine Learning and Knowledge Discovery in Databases. Research Track | European Conference, ECML PKDD 2024, Vilnius, Lithuania, September 9-13, 2024, Proceedings, Part II
Brand: Springer
Price: 85.59 EUR
Availability: OnlineOnly

European Conference, ECML PKDD 2024, Vilnius, Lithuania, September 9-13, 2024, Proceedings, Part II

Albert Bifet Jesse Davis Tomas Krilavicius Meelis Kull Eirini Ntoutsi Indre Zliobaite(Editor)

Springer (Publisher)

Published on 29. August 2024

LVII, 462 pages

E-Book

PDF with digital watermarking

System requirements

978-3-031-70344-7 (ISBN)

€85.59incl. 7% vat

System requirements

for PDF with digital watermarking

E-Book Single Licence

Available for download

Description

Alles über E-Books | Antworten auf Fragen rund um E-Books, Kopierschutz und Dateiformate finden Sie in unserem Info- & Hilfebereich.

Alles über E-Books, Kopierschutz & Dateiformate finden Sie in unserem Info- & Hilfebereich.

This multi-volume set, LNAI 14941 to LNAI 14950, constitutes the refereed proceedings of the European Conference on Machine Learning and Knowledge Discovery in Databases, ECML PKDD 2024, held in Vilnius, Lithuania, in September 2024.

The papers presented in these proceedings are from the following three conference tracks: -

Research Track: The 202 full papers presented here, from this track, were carefully reviewed and selected from 826 submissions. These papers are present in the following volumes: Part I, II, III, IV, V, VI, VII, VIII.

Demo Track: The 14 papers presented here, from this track, were selected from 30 submissions. These papers are present in the following volume: Part VIII.

Applied Data Science Track: The 56 full papers presented here, from this track, were carefully reviewed and selected from 224 submissions. These papers are present in the following volumes: Part IX and Part X.

More details

Other editions

Content

Intro
Preface
Organization
Invited Talks Abstracts
The Dynamics of Memorization and Unlearning
The Emerging Science of Benchmarks
Enhancing User Experience with AI-Powered Search and Recommendations at Spotify
How to Utilize (and Generate) Player Tracking Data in Sport
Resource-Aware Machine Learning-A User-Oriented Approach
Contents - Part II
Research Track
Landscape Analysis of Stochastic Policy Gradient Methods
1 Introduction
2 Preliminaries
3 Uniform Convergence Results
3.1 Uniform Convergence of Objective
3.2 Uniform Convergence of Gradient
3.3 Uniform Convergence of Hessian
3.4 Proof Sketch
4 Characterizing the Landscape of Policy Gradient Methods
5 Conclusion
References
Spatiotemporal Covariance Neural Networks
1 Introduction
2 Related Works
3 Problem Formulation
4 Spatiotemporal Covariance Neural Networks
5 Stability Analysis
5.1 Stability of Spatiotemporal Covariance Filter
5.2 Stability of STVNN
6 Numerical Results
6.1 Model Analysis
6.2 Forecasting
7 Conclusion
References
Frequency Enhanced Pre-training for Cross-City Few-shot Traffic Forecasting
1 Introduction
2 Related Work
3 Preliminary
4 Method
4.1 Pre-training
4.2 Fine-tuning
5 Experiment
5.1 Experiment Settings
6 RQ1: Overall Performance
7 RQ2: Ablation Study
8 RQ3: Reconstruction Analysis
9 RQ4: Cross Domain Analysis
10 Conclusion
References
Simple Graph Condensation
1 Introduction
2 Related Work
3 Problem Definition
4 Method
4.1 The Condensed Graph
4.2 The Pre-trained SGC
4.3 Representation Alignment
4.4 Logit Alignment
4.5 Optimization
5 Experiments
5.1 Experiment Settings
5.2 Prediction Accuracy
5.3 Condensation Time
5.4 Generalizability Capability
5.5 Neural Architecture Search
5.6 Knowledge Distillation
5.7 Statistics of Condensed Graphs
5.8 Ablation Study
6 Conclusion
References
Multivariate Traffic Demand Prediction via 2D Spectral Learning and Global Spatial Optimization
1 Introduction
2 Related Work
2.1 Spectral Representation Learning
2.2 Traffic Demand Forecasting
3 Methodology
3.1 Problem Definition
3.2 Embedded 2D Spectral Learning
3.3 Global Spatial Optimization
3.4 Overall Architecture and Loss Function
4 Experiment
4.1 Datasets
4.2 Baselines and Evaluation Metrics
4.3 Implementation Details
4.4 Comparison with the State-of-the-Art
4.5 Ablation Study
4.6 Multi-horizon Forecasting Comparison
4.7 Frequency Component Visualization
4.8 Parameter Analysis
5 Conclusion
References
Reliable Classifications with Guaranteed Confidence Using the Dempster-Shafer Theory of Evidence
1 Introduction
2 Related Work
3 Preliminaries
3.1 Dempster-Shafer Theory of Evidence
3.2 Conformal Prediction
4 Method
5 Numerical Experiments
6 Conclusion
References
LimGen: Probing the LLMs for Generating Suggestive Limitations of Research Papers
1 Introduction
2 Related Work
3 LimGen Dataset
3.1 Dataset Collection
3.2 Nature of the Limitations
4 Benchmark Experiments
4.1 Task Formulation
4.2 Methodology
4.3 Experimental Setup
5 Experimental Results and Analysis
5.1 Evaluation
6 Challenges and Future Work
7 Limitations
8 Conclusions
References
On the Robustness of Global Feature Effect Explanations
1 Introduction
2 Related Work
2.1 Global Feature Effect Explanations
2.2 Robustness and Stability of Explanations
3 Notation and Definition of Feature Effects
4 Theoretical Analysis
4.1 Robustness to Data Perturbation
4.2 Robustness to Model Perturbation
5 Experiments
5.1 Robustness to Data Perturbation
5.2 Robustness to Model Perturbation
6 Conclusion
A Experiments: Robustness to Data Perturbation
B Experiments: Robustness to Model Perturbation
References
Federated Learning with Flexible Architectures
1 Introduction
2 Related Work
2.1 Heterogeneous Network Aggregation in Federated Learning
2.2 Split Learning
2.3 Skip Connections
3 Motivation: Challenges in Heterogeneous Aggregation
3.1 Security Concerns of Heterogeneous Network Aggregation
3.2 Scale Variations in Heterogeneous Networks
4 Flexible Federated Learning
4.1 FedFA Procedure
4.2 Layer Grafting Method for Ensuring Security
4.3 Scalable Aggregation: Normalization of Local Model Weights
4.4 Global Model Distribution Step
5 Experimental Evaluation
5.1 Experimental Setup
5.2 Baselines and Metrics
5.3 Evaluation
6 Future Work and Limitations
7 Conclusion
References
A Unified Data Augmentation Framework for Low-Resource Multi-domain Dialogue Generation
1 Introduction
2 Methodology
2.1 Problem Formulation
2.2 De-domaining Data Processing
2.3 Domain-Agnostic Training and Domain Adaptation
2.4 Domain Similarity
3 Experiments
3.1 Datasets
3.2 Models
3.3 Baselines
3.4 Implementation Details
3.5 Evaluation Metrics
4 Results and Analysis
4.1 Overall Results
4.2 Impact of Domain Similarity
4.3 Impact of Dataset Size
4.4 Human Evaluation Results
5 Related Work
6 Conclusion
References
Improving Diversity in Black-Box Few-Shot Knowledge Distillation
1 Introduction
2 Related Works
3 The Proposed Framework
3.1 Problem Statement
3.2 Proposed Method
4 Experiments
4.1 Architectures and Datasets
4.2 Baselines
4.3 Distillation Performance
4.4 Diversity of Our Synthetic Images
4.5 Ablation Studies
4.6 Comparison with Data-Free KD Methods
5 Conclusion
References
Self-pro: A Self-prompt and Tuning Framework for Graph Neural Networks
1 Introduction
2 Related Work
2.1 Graph Neural Networks
2.2 Graph Pre-training
2.3 Graph Prompt Learning
3 Preliminaries
4 Methods
4.1 Pre-training Method
4.2 Prompt Template
4.3 Self-prompt Architecture
4.4 Self-prompt Injection and Tuning
5 Experiments
5.1 Experimental Settings
5.2 Few-Shot Node Classification (RQ1)
5.3 Performance Under Different Shots and Tasks (RQ2)
5.4 Ablation Study (RQ3)
5.5 Parameter Analysis (RQ4)
6 Conclusion
References
Variable-Agnostic Causal Exploration for Reinforcement Learning
1 Introduction
2 Related Work
3 Methods
3.1 Background
3.2 Variable-Agnostic Causal Exploration Reinforcement Learning Framework
4 Experiments
4.1 VACERL: Causal Intrinsic Rewards Implementation and Evaluation
4.2 VACERL: Causal Subgoals Implementation and Evaluation
4.3 Ablation Study and Model Analysis
5 Conclusion
References
LayerGLAT: A Flexible Non-autoregressive Transformer for Single-Pass and Multi-pass Prediction
1 Introduction
2 Methodology
2.1 Non-autoregressive Generative Models
2.2 GLAT and CMLM Training Strategies
2.3 The Proposed Model: LayerGLAT
3 Experimental Set-Up
3.1 Datasets
3.2 Models and Settings
3.3 Knowledge Distillation
3.4 Noisy Parallel Decoding
4 Main Results
4.1 Performance of Single-Pass Generation
4.2 Iterative Performance
4.3 Inference Speed-Up
5 Ablation and Sensitivity Analysis
5.1 Performance With Reduced Model Size
5.2 Sampling Rate
5.3 Sampling Strategy
5.4 Performance with Raw Data Training
6 Related Work
7 Conclusion
A Dataset-Specific Hyperparameters
B Training Algorithm
References
Generative Modeling with Flow-Guided Density Ratio Learning
1 Introduction
2 Background
2.1 Wasserstein Gradient Flows
2.2 Density Ratio Estimation via Bregman Divergence
3 Discriminator Gradient Flow
3.1 Where The Stale Estimator Breaks Down
3.2 Data-Dependent Priors
4 Flow-Guided Density Ratio Learning
4.1 Convergence Distribution of FDRL
5 Related Works
6 Experiments
6.1 Setup
6.2 Unconditional Image Generation
6.3 Ablations
6.4 Conditional Image Generation
6.5 Unpaired Image-to-Image Translation
7 Conclusion
References
ADR: An Adversarial Approach to Learn Decomposed Representations for Causal Inference
1 Introduction
2 Notations and Problem Setup
3 Theoretical Analyses
3.1 Motivation: Variance Bound for the CATE
3.2 Definitions and Theoretical Analyses of I, C, A
4 ADR Algorithm
4.1 Overview
4.2 Loss Functions for Decomposed Representations
4.3 Adversarial Learning of Decomposed Representations
5 Experiments
5.1 Compared Methods and Evaluation Metrics
5.2 Experiments on Synthetic Datasets
5.3 Experiments on Real Datasets
6 Conclusion and Discussion
References
Data is Moody: Discovering Data Modification Rules from Process Event Logs
1 Introduction
2 Related Work
3 Preliminaries
3.1 Notation for Data Modification Rules
3.2 Minimum Description Length
4 MDL for Data Modifications
4.1 Data Encoding
4.2 Model Encoding
4.3 Formal Problem Definition
5 The Moody Algorithm
5.1 Estimating the MDL Score
5.2 Finding Good Modification Rules
6 Experiments
6.1 Synthetic Event Logs
6.2 Real-World Event Logs
7 Discussion
8 Conclusion
References
Self-certified Tuple-Wise Deep Learning
1 Introduction and Related Work
1.1 Related Work
2 Preliminaries
2.1 Tuple-Wise Learning via U-Statistics
3 Main Results
3.1 Optimization of the Bound w.r.t. Posterior Q
3.2 Plugging the Obtained Q Into Theorem 1
4 Experiments with Application to Metric Learning
4.1 Dataset and Model
4.2 PAC-Bayes Bound Minimization Algorithm
4.3 Results
5 Conclusions
A Proofs
References
How Much Training Data Is Memorized in Overparameterized Autoencoders? An Inverse Problem Perspective on Memorization Evaluation
1 Introduction
2 Recovery of Training Data as an Inverse Problem
2.1 Overparameterized Autoencoders: A Definition
2.2 The Recovery Problem
2.3 The Inverse Problem Perspective to Training Data Recovery
3 A Practical Algorithm for Training Data Recovery
3.1 Estimation of a Training Sample x for a Given Degradation Operator Estimate
3.2 Estimation of a Degradation Operator for a Given Training Sample Estimate
4 Experimental Results
5 Conclusion
A Proof of Theorem 1
B Proof of Equation (17)
C The Examined Autoencoder Architectures
C.1 Perfect Fitting Regime
C.2 Moderate Overfitting Regime
D The Proposed Method: Additional Implementation Details
D.1 Stopping Criterion of the Proposed Method
D.2 Gamma Values for the Proposed Method
References
Continual Neural Computation
1 Introduction
2 Neural Computation
2.1 Insights on the Computational Scheme
3 Continual Learning from a Stream of Data
4 Related Work
5 Experiments
5.1 Results
6 Conclusions and Future Work
References
PINN-BO: A Black-Box Optimization Algorithm Using Physics-Informed Neural Networks
1 Introduction
2 Related Works
3 Preliminaries
4 Proposed Method
5 Theoretical Analysis
6 Experimental Results
6.1 Baselines
6.2 Synthetic Benchmark Functions
6.3 Real-World Applications
7 Conclusion
References
Enhancing Sharpness-Aware Minimization by Learning Perturbation Radius
1 Introduction
2 Related Work
3 The LETS Method
3.1 Problem Formulation
3.2 Learning Perturbation Radius for SAM
3.3 LETS-ASAM
4 Experiments
4.1 CIFAR-10 and CIFAR-100
4.2 ImageNet
4.3 IWSLT'14 DE-EN
4.4 GLUE
4.5 Robustness to Label Noise
4.6 Robustness to the Initialization of
4.7 Loss Landscapes
4.8 Effects of Generalization Metrics
4.9 Convergence
5 Conclusion
References
SaccadeDet: A Novel Dual-Stage Architecture for Rapid and Accurate Detection in Gigapixel Images
1 Introduction
2 Related Work
3 Proposed Method
3.1 Preliminaries
3.2 Overall Architecture
3.3 Saccade by Multi-scale Density Estimation
3.4 Scale-Aware Mean Squared Error Loss
3.5 Patch Generation
3.6 Gaze with Scale Normalization
4 Experiments
4.1 Baseline
4.2 Module Analysis
4.3 Comparisons with the State-of-the-Art Methods
4.4 Ablation Studies
4.5 Application to Whole Slide Imaging
5 Conclusion
References
Physics-Informed Spatio-Temporal Model for Human Mobility Prediction
1 Introduction
2 Related Work
3 Preliminaries
4 Methodology
4.1 Physics-Informed Human Mobility Energy Equation
4.2 Neural Networks for PIE
4.3 Model Training
5 Experiments
5.1 Experimental Setting
5.2 Comparison with Baselines (RQ1)
5.3 Ablation Study and Analysis (RQ2)
5.4 Case Study (RQ3)
5.5 Qualitative Evaluation (RQ4)
6 Conclusion
References
Adaptive Seasonal-Trend Decomposition for Streaming Time Series Data with Transitions and Fluctuations in Seasonality
1 Introduction
2 Related Work
2.1 Seasonal-Trend Decomposition (STD)
2.2 Season Length Estimation (SLE)
3 Preliminaries and Problem Setting
4 Online Season Length Estimation
4.1 Periodogram
4.2 Sliding Discrete Fourier Transform (SDFT)
4.3 Spectral Peak Location Estimation
5 Algorithm
6 Experimental Settings
7 Experimental Results
7.1 Experimental Results with Synthetic Datasets
7.2 Experimental Results with Real-World Datasets
8 Conclusion
References
A Deep Cut Into Split Federated Self-Supervised Learning
1 Introduction
2 Related Work
3 Preliminaries
3.1 Federated Self-Supervised Learning
3.2 Momentum Contrastive Split Federated Learning
4 MonAcoSFL: Momentum-Aligned Contrastive Split Federated Learning
4.1 Limitations of MocoSFL
4.2 MonAcoSFL
5 Experimental Investigation
5.1 Experimental Setup
5.2 Accuracy Performance
5.3 Privacy Evaluation
5.4 Analysis of Models Alignment
6 Conclusion
References
Author Index

System requirements

Save as PDF Copy link into clipboard

Schweitzer Fachinformationen

Machine Learning and Knowledge Discovery in Databases. Research Track

Description

More details

Other editions

Additional editions

Content

System requirements