Master Seminar "Machine Learning with Knowledge Graphs" (WS 2023/24)
- Contact: Prof. Dr. Volker Tresp
- Required: Lecture "Machine Learning" or equivalent.
Registration: via Moodle (central allocation)
Termine und Ort
|Do, 16:00-18:00||Amalienstr. 73A, 101||19.10.2023|
Knowledge graphs (KGs) describe structured symbolic data by representing information via entities and their relationships. This form of relational knowledge representation has a long history in logic and artificial intelligence.
Relational embedding methods are useful methods for representing KGs and an important step towards KG reasonings, including link prediction in AI systems. For example, RESCAL, which was the first embedding model for KGs, has been used for recommending industrial components to customers.
At the same time, relationships between entities, in reality, are evolving, and temporal knowledge graphs are also of great significance in this respect. Events in temporal KG (tKG) are described as quadruple with an additional timestamp, for example, (Obama, is, USA president, 2008-2016).
With pre-trained foundation models like GPT-4, knowledge graphs even gain more attention to help reduce hallucination, increasing the factuality and reasoning ability of black-box large models.
The seminar focuses on the above-mentioned topics of machine learning with Knowledge Graphs, and students are encouraged to investigate state-of-the-art work in this field to learn cutting-edge knowledge graph research.
 Nickel, Maximilian, Volker Tresp, and Hans-Peter Kriegel. "A three-way model for collective learning on multi-relational data." *Icml*. Vol. 11. No. 10.5555. 2011.
 Adesso, Gerardo. "GPT4: The ultimate brain." *Authorea Preprints* (2022).