Open Topics
In the context of probabilistic inference
- Scalability of the colouring algorithm for lifting models and adaptation of noisy symmetries (BA / MA)
- Applying lifting to tensor networks (BA / MA)
- Sampling in the (lifted) junction tree algorithm
- Lifting of variable eliminination in the Fourier domain (MA)
- Lifting importance sampling in parfactor graphs (MA)
In the context of decision making
- A comparative analysis of POSG (partially observable stochastic game) solution methods
In the context of security & privacy
- Analysis of spectral clustering for differential privacy (more MA level)
- Oblivious databases (BA/MA; specific focus depends on personal interests; can range from a literature study or a formal analysis to implementation; in collaboration with Thore Thießen, RG Efficient Algorithms)
In the context of text modelling
-
Challenges of dynamic topic modelling given small domain-specific corpora (BA)
-
Learning relational models using RDF triples extracted from text (BA / MA)
Hardware-dependent
- Implementing and evluating methods of probabilistic inference using GPUs
In collaboration with the Institute of Medical Informatics (Tobias Brix)
- Development of plugins for the Docu-Wiki (BA)
- Database-related topics with different foci can be arranged