Local Structure Discovery in Bayesian Networks

TITLE
Local Structure Discovery in Bayesian Networks

ABSTRACT
Learning a Bayesian network structure from data is an NP-hard problem and thus exact algorithms are feasible only for small data sets. If a modeler is interested in only few target variables and the local structure near them, it is possible to scale up by using local learning. We present a score-based local learning algorithm called SLL. Empirical results suggest that SLL is competitive when compared to the constraint-based HITON algorithm. We also study the prospects of learning the global structure based on local results.

ABOUT THE PRESENTER
Teppo Niinimäki is a doctoral student at the University of Helsinki, Department of Computer Science.