Studying Scientific Innovation with Temporal Knowledge Graph Representation Learning
Description
What’s the next big idea, and who’s going to discover it? Our project is trying to understand how researchers make new discoveries and innovate new ideas. To do that we will apply deep learning techniques for temporal knowledge graph learning (RE-NET, CyGNet, HINGE, StarE) to a huge citation network dataset. We have assembled a KG with 260M research papers, 270M authors, 700K fields. To learn representations, our training tasks include citation prediction, author collaboration prediction, and field of study prediction.
Advisors
Skills Required by the team
- Python
- Knowledge Graphs
- PyTorch
- Tensorflow
- Wikidata