Abstract

Abstract. Sakama and Inoue introduced brave induction as a novel logic framework for concept-learning. They showed that brave induction has potential applications for problem solving in many domains. In this paper, motivated from Shapiro's definition of model inference problems, we provide an optimization of brave induction called proper brave induction, which prefers hypotheses resulting fewer minimal models. We first propose formal definitions of proper brave induction for clausal theories and nonmonotonic logic programs, then investigate corresponding properties and develop an optimization procedure. At last, we analyze computational complexity of decision problems for proper brave induction in propositional case.