Masked autoencoders for contrastive learning of heterogenous graphs
In this data driven society, information networks are mostly heterogenous which consists of different types of entities and relationships. Heterogenous Graph Neural Networks utilize Heterogenous graphs to study and understand the data. Since Heterogenous Graph Neural Network uses semi-supervised lea...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | |
| Format: | Final Year Project |
| Language: | English |
| Published: |
Nanyang Technological University
2024
|
| Subjects: | |
| Online Access: | https://hdl.handle.net/10356/176838 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Institution: | Nanyang Technological University |
| Language: | English |
| Summary: | In this data driven society, information networks are mostly heterogenous which consists of different types of entities and relationships. Heterogenous Graph Neural Networks utilize Heterogenous graphs to study and understand the data. Since Heterogenous Graph Neural Network uses semi-supervised learning or supervised learning, it is not ideal for real-life scenarios and therefore, certain part of the data will be masked when training. Generating data specific models makes studying these models relatively easier. This allows model to be more robust, capable to work for various datasets. Therefore, it is ideal to receive ideal accuracy for this model to verify the working condition of this model. Modifying various parameters allows the accuracy to differ and knowing the right parameters is ideal. |
|---|