Test set confusion matrix for SED type classification. Each row of the matrix represents the fraction of samples in an actual SED type, while each column represents the fraction of samples in a predicted SED type. The superimposed numbers indicate the fraction of objects of a class in that particular position. For example, considering the second row (SBs): 96% of SBs are correctly classified as such, 3% as SB-AGN, and 1% as Irregular.