Advanced Probabilistic Systems

3 credits

Advanced models and algorithms for tractable probabilistic inference, with emphasis on circuit-based representations of probability distributions. Topics include probabilistic circuits such as Probabilistic Sentential Decision Diagrams (PSDDs), arithmetic circuits, Noisy-OR Bayesian Networks and related decomposable and deterministic representations. The course covers compilation of graphical models into circuits, structural properties enabling efficient inference (decomposability, determinism, smoothness), and complexity guarantees.