Artificial datasets: example patterns, with the true underlying parcellations used to generate the data.

<p>The inter-individual variability is controlled by two parameters: i) the locations of the middle parcels, which induces their overlap, here 33% (variability in geometric attributes); and ii) the value of , which induces the amount of variation in the mean intensities observed in the parcels (variability in activation attributes). Although the parcels are matched across subjects by construction, the colors of the true parcels are chosen different in the two subjects to illustrate that G-SVC does not need an a priori match.