Class structure recovered by non-negative tensor factorization as a function of the number of components.

Each row of the table corresponds to one tensor decomposition, with a number of components ranging from to . For each decomposition, we report the core consistency, the number of classes that can be matched to the extracted components, the total number of nodes (students) these classes are known to comprise, the number of nodes that belong to the component (i.e., the class nodes covered by the component), and the ratio between the latter two columns, that is, the fraction of the known class structure that the method has recalled.