BayesStore: Probabilistic Graphical Models as First Class Database Citizens

There is an increasing demand for managing and reasoning about probabilistic data including for example noisy sensor data arising from ubiquitous computing. On top this data often includes complex correlation patterns. Currently Probabilistic Database Systems support uncertainty usually at individual tuple or attribute level allowing for fine-grained uncertainty but often also resulting in unnecessary complex query processing. In addition, more complex reasoning or computation is performed outside of the database, hence the full data set needs to be extracted from the database, compiled into an input file, then processed by an external tool and the results written back to the database.