Rodrigo Basilio,  Gerson Zaverucha,  Valmir Carneiro Barbosa. 

First-order theory refinement using neural networks is still an o pen problem. Towards a solution to this problem, we define a First-Order ext ensio n of the Cascade ARTMAP (FOCA) system, using Inductive Logic Pro gramming techniques. We: a) modify the network structure to handle first -o rder objects; b) define first-order versions of the choice (similarity) funct io n and of the vigilance criterion, the main functions that guide all Cascade ARTMAP dynamics; c) define a first-order version of the pro po sit io nal learning algorithm, that approximates Plotkin's least general generalizat io n (lgg). Results show that our initial goal, learning logic pro grams using neural networks, has been achieved. Clique no link abaixo para buscar o texto completo deste trabalho na Web: Buscar na Web Processando consulta.