Reinforcement Learning for Knowledge Graphs
The development of knowledge graphs has increased significantly recently. However, it is very challenging to efficiently develop and maintain large-scale knowledge graphs. In this thesis, you will use classic models of reinforcement learning to target two tasks in knowledge graphs.
We offer two topics to use reinforcement learning on knowledge graphs:
(1) Link prediction: for example, predict how likely there exists a relation between two entities.
(2) Entity summarization: extract top-k most relevant triples that relate to users' interest.
It would be good if you have taken or plan to take the following courses:
- IN3060/IN4060 Semantic technologies
- IN3070/IN4070 Logikk
- IN-STK5100 Reinforcement Learning and Decision Making Under Uncertainty
- IN-STK5000 Adaptive methods for data-based decision making