Preliminary Content
Acknowledgements
Preface
Dedication
Abstract
1
Introduction
1.1
Motivation
1.2
Research Objectives
2
Background
2.1
Preliminaries
2.2
Applications of Machine Learning on Graphs
2.2.1
Node-level Tasks
2.2.2
Set-level Tasks
2.2.3
Graph-level Tasks
2.3
Graph Representation Learning
2.3.1
Random Walk and Factorization-based Methods
2.3.2
Graph Neural Networks
2.3.3
Expressiveness of Graph Neural Networks
2.3.4
Design Space of Graph Neural Networks
2.4
Link Prediction
2.4.1
Learning Representations for Link Prediction
2.4.2
Inductive Link Prediction
3
Design Space for Inductive Link Prediction
3.1
Standardized Architecture
3.2
Proposed Design Space
3.3
Experimental Design
3.3.1
Evaluation of Model Generalization
3.3.2
Identification of Condensed Design Space
3.3.3
Quantification of Task Difficulty
4
Evaluation
4.1
Methodology
4.2
Evaluation of Model Generalization
4.2.1
Aggregate analysis
4.2.2
Performance by Design Dimension
4.2.3
Dimension-level effects
4.3
Reduced Design Space
4.4
Task Difficulty
5
Software Developed
Conclusion
Appendix
A
The First Appendix
B
The Second Appendix, for Fun
References
Published with bookdown
Design Space Exploration of Graph Neural Networks for Inductive Link Prediction
Chapter 5
Software Developed