Overlapping between semantic categories and D-MST neuronal-based clusters.

The table reports the overlap (fifth column) between each semantic category (first column) and the D-MST neuronal-based cluster (second column) containing the best matching sub-tree of contiguous objects, according to a score defined as the ratio between the intersection of the sub-tree with the category and their union (fifth column). Significance of the overlap was computed by permuting (1,000,000 times) either sets of twin objects (forth- and third-to-last columns) or individual objects (second-to-last and last columns) across the categories of a given clustering hypotheses: Holm-Bonferroni corrected p<0.01 (**) and p<0.05 (* and *+); and uncorrected p<0.01 (++ and *+) and p<0.05 (+). For comparison with [14], two other overlap metrics (Ratio 1 = the fraction of objects in the category overlapping with the cluster; and Ratio 2 = the fraction of objects in the cluster overlapping with the category) are also reported.