Examining and comparing two approaches to data training in descriptive-separative dictionary

In this article, we examine the two main approaches proposed in the field of dictionary learning. The first approach is aimed at creating separation in the representations provided by the dictionary, that is, all atoms are simultaneously descriptive and separable. The second approach is to create a subspace for the descriptiveness of the dictionary and create another subspace for its differentiation. The classification in the first approach is on the full display area, and the classification in the second approach will be done only with the part of the display that corresponds to the separating atoms.