Hi Nobuko,

(1) In slow_surf_clustsim.py I use rmm=-1, which is just like
using nearest neighbor. That seems simple and reasonable, and
it is just for forming clusters. To use an rmm value that is
bigger means allowing for clusters to stretch over small values.

(2) Volume based comes from mapping EPI data onto the surface,
for example. Since the initial voxels are large, relative to
the node density, clustering will tend to change in chunks.

The "on the surface" example might be based on something like
cortical thickness measurements, which effectively start on
the surface (though one could consider using an anatomical
volume as the basis).

The point is to try to mimic the actual flow of the data.

- rick