able in common with the previous relations. Otherwise the patterns would combine many unrelated things. In frequent itemset mining (for example the APRIOR] algorithm), there is effectively one variable, and adding a new predicate will often decrease the number of items that match. It can never increas