Validation macro recalls of the random forest classifiers trained on the augmented representations ${\stackrel{\sim }{\mathbf{C}}}^{}\mathbf{\text{DC+V}}$, ${\stackrel{\sim }{\mathbf{C}}}^{}\mathbf{\text{SD+V}}$, ${\stackrel{\sim }{\mathbf{C}}}^{}\mathbf{\text{C+V}}$, and ${\stackrel{\sim }{\mathbf{C}}}^{}\mathbf{\text{F+V}}$ of the synthetic samples W0C0 (a), W100C0 (b), W0C10L50 (c), and W100C10L50 (d), evaluated as a function of the number of coefficients in the representation. The errorbars show the minimum and the maximum validation macro recalls for a given number of coefficients. The solid black lines represent the best validation macro recalls achieved by the classifiers trained on extended full representations of the samples. We describe the synthetic samples in Sect. 2 and provide details about the augmented representations in Sect. 3.4.