Temporal Properties of Conditional Independence in Dynamic Bayesian Networks
From International Center for Computational Logic
Temporal Properties of Conditional Independence in Dynamic Bayesian Networks
Talk by Rajab Aghamov
- Location: APB 3027
- Start: 28. January 2027 at 11:00 am
- End: 28. January 2027 at 12:00 pm
- Research group: Knowledge-Based Systems
- Event series: Research Seminar Logic and AI
- iCal
Dynamic Bayesian networks (DBNs) provide compact graphical representations of probabilistic systems that evolve over time. This talk presents results on verifying the evolution of conditional-independence propositions against temporal specifications. Two interpretations are considered: stochastic conditional independence, which takes the concrete probability distributions into account, and structural conditional independence, which depends only on the graphical structure. Deciding whether a stochastic conditional-independence proposition eventually holds is shown to be Skolem-hard. For structural conditional independence, verification against linear temporal logic (LTL) specifications and NBA is in PSPACE and is both NP-hard and coNP-hard.
