Learning a mixture of graph neural networks
GNN models are designed to handle complex and non-uniform graph-structured data for classification tasks such as node, link, and graph-level predictions. However, the randomness and unfixed node ordering of graph data still proposes a challenge to GNN models. To improve the accuracy of GNN mod...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2023
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/167150 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | GNN models are designed to handle complex and non-uniform graph-structured data for
classification tasks such as node, link, and graph-level predictions. However, the randomness
and unfixed node ordering of graph data still proposes a challenge to GNN models. To improve
the accuracy of GNN models, this study proposes a mixture model approach that combines
multiple renowned GNN architectures to produce a single output. The study discusses the
advantages of several GNN architectures and tests the models on selected benchmark datasets to
examine their validity. By comparing simple GNN models to mixture models, the study hopes to
provide insights into the development of more accurate GNN models. |
|---|