In this paper, we present a novel unsupervised learning approach inspired by cognitive models. Unlike previous unsupervised learning methods that mainly focus on clustering samples in a mathematical space, in this paper, we propose a primitive-based unsupervised learning method that constructively models the input space as a distributed hierarchy independent of the input data. We demonstrate the superiority of the proposed method by comparing it with the existing state-of-the-art models in various fields such as state-of-the-art unsupervised learning classification, classification of small and incomplete datasets, and cancer type classification, and show that the proposed method outperforms existing algorithms (including supervised learning) and exhibits cognitive-like behavior through cognitive feature evaluation.
