Probabilistic Foundation Models

Organisation

• Lecture (Prof. Dr. Ralf Möller, Dr. Marcel Gehrke)
• Seminar (Dr. Marcel Gehrke)

Content

• Parametric factor graphs (PFGs), lifted Inference: lifted variable elimination, lifted branching tree algorithm, model counting (first order and algebraic type), relational probabilistic computational networks
• Sequential (e.g. discrete-time) modeling and inference with PFGs, taming of PFGs over time (retrospective and progressive) discrete-time) modeling and inference with PFGs, taming PFGs over time (retrospective and progressive)
• Machine learning for PFGs
• Decision making and planning with PFGs and under causality considerations
• Dynamic extensions of the state space: Generative dynamic causal probabilistic-relational models for stochastic games (genDC-SG-PFGs)
• PFGs and LLMs