Concept-cognitive learning is an emerging cross-cutting field that learns new knowledge by modeling the cognitive processes of the human brain. However, the existing concept-cognitive learning models have ignored the problems of the fit between the extent and intent of concepts, the redundancy and structural characteristics of concept space as well. To address these challenges, this paper introduces a novel cognitive learning model grounded in the concept of approximation. Firstly, incomplete concepts are generated by constructing attribute-object λ operator and thus the fit between the extent and intent of concepts is improved. Aiming at the redundancy of concept space, further discussion is conducted on the set of upper approximation concepts, the collection of lower approximation concepts, and the family of boundary concepts. Subsequently, by taking into account the structural characteristics of concept space, the upper and lower approximation concept spaces are established respectively, and a novel learning accuracy is then proposed to portray the cognitive learning effect. Finally, the effectiveness of the proposed method is verified by simulation experiments.

Concept cognitive learning (CCL) constitutes a rigorous and current cognitive theory for the representation and learning of concepts of the human brain. The well-established CCL models pay close attention to construct a concept space, in which all the attributes are mastered concurrently. However, few attempts have been made to combine CCL with cognitive logic in a fuzzy context due to attribute precedence. For this case, this paper first develops a cognitive surmise relationship among attributes, cognitive transitions, and discrimination of fuzzy concepts based on knowledge space theory. Furthermore, utilizing the inherent information of concepts, the attributes are weighted to accurately understand and apply fuzzy concepts. To better derive benefits from the fuzziness and uncertainty of knowledge, an approach is provided to improve performance through the fusion of fuzzy concepts. Empirical studies on twenty datasets reveal the effectiveness and efficiency of the proposed model.