Olivier Devillers a écrit : > > > Just to let you know > > > I add some details in the computations (but a lot of things has to > be polished) > > I justified the point raised by Xavier (conditioning of the size of > the big region by the fact that the small ones are non empty) > > I made some computation in higher dimension (where the constraint on > delat in terms of n are stronger) > > > I do nat have a clean explanation for the fact that for delta =1 we > get a better result that > for points random in a sphere. (even in 2d) > Either - we miss some constrain on delta forbiding to take delta = 1 > - there is a mistake somewhere > - we are right : the convex hull size is first decreasing > then increasing when delta increase ? explanation added + experiments
Attachment:
convex-hull.pdf
Description: convex-hull.pdf