RELATIONAL-LEARNING REEXAMINED - AUTHORS RESPONSE

We argue that existing learning algorithms are often poorly equipped t o solve problems involving a certain type of important and widespread regularity that we call ''type-2 regularity''. The solution in these c ases is to trade achieved representation against computational search. We investigate several ways in which such a trade-off may be pursued including simple incremental learning, modular connectionism, and the developmental hypothesis of ''representational redescription''.