Conclusions
	n Framework for active learning for structure in BNs n Theoretical gains supported by experimental results n Efficient algorithm n Key idea: n obtain measure of model quality n choose query that maximizes the quality n Future work: missing data, continuous variables