— Emerging uncertain database applications often involve the cleansing (conditioning) of uncertain databases using additional information as new evidence for reducing the uncertainty. However, past researches on conditioning probabilistic databases, unfortunately, only focus on functional dependency. In real world applications, most additional information on uncertain data sets can be acquired in the form of aggregate constraints (e.g., the aggregate results are published online for various statistical purposes). Therefore, if these aggregate constraints can be taken into account, uncertainty in data sets can be largely reduced. However, finding a practical method to exploit aggregate constraints to decrease uncertainty is a very challenging problem. In this paper, we present three approaches to cleanse (condition) uncertain databases by employing aggregate constraints. Because the problem is NP-hard, we focus on the two approximation strategies by modeling the problem as a nonlinea...