Title Page
Abstract
Contents
Chapter 1. Introduction 9
Chapter 2. Related Work 11
2.1. Graph Neural Networks 11
2.2. Graph Collaborative Filtering 11
2.3. Self-supervised Learning 12
Chapter 3. Methodology 14
3.1. Overview 14
3.2. Problem Definition 14
3.3. Graph Denoising Module 15
3.4. Self-supervised Learning Module 17
3.5. Prediction 19
Chapter 4. Experiments 20
4.1. Datasets 20
4.2. Baselines 21
4.3. Evaluation Metrics 21
4.4. Implementation Details 22
4.5. Overall Performance 22
4.6. Ablation Study 26
Chapter 5. Conclusion 29
Bibliography 30
초록 35
Table 1. Statistics of the datasets. 20
Table 2. The overall performance comparison on the CiteULike dataset. The best result is bolded and the runner-up is underlined. 23
Table 3. The overall performance comparison on the Movielens-1M dataset. The best result is bolded and the runner-up is underlined. 24
Table 4. The overall performance comparison on the LastFM dataset. The best result is bolded and the runner-up is underlined. 25
Figure 1. An illustration of RBS model architecture. 15
Figure 2. Ablation study on CiteULike dataset - Precision. 27
Figure 3. Ablation study on CiteULike dataset - Recall. 27
Figure 4. Ablation study on CiteULike dataset - NDCG. 28