Most current applications of inductive learning in databases take plac
e in the context of a single extensional relation. This paper puts ind
uctive learning in the context of a set of relations defined either ex
tensionally or intentionally in the framework of deductive databases.
It presents LINUS, an inductive logic programming system that induces
virtual relations from example positive and negative tuples and alread
y defined relations in a deductive database. Based on the idea of tran
sforming the problem of learning relations to attribute-value form, it
incorporates several attribute-value learning systems. As the latter
handle noisy data successfully, LINUS is able to learn relations from
real life noisy databases. The paper illustrates the use of LINUS for
learning virtual relations and then presents a study of its performanc
e on noisy data.