Fig. 6: Example of prototype and criticism explanations. | npj Computational Materials

Fig. 6: Example of prototype and criticism explanations.

From: Explainable machine learning in materials science

Fig. 6

Kim et al.62 found random subsets of prototypes and criticisms of the USPS handwritten digits data set157 by applying the MMD-critic technique62. Prototypes and criticisms are interesting because prototypes can benefit data understanding while criticisms can benefit data cleaning. Figure reprinted from ref. 62 with permission.

Back to article page