Efficient learning algorithm and the theoretical foundations for concept learning, which is a principle aspect of computational learning theory, are investigated. Especially, we focus on the approach of using monotone concepts to approximating arbitrary concepts.