But you have to be careful about that word "independent".

There's a reason that things like 3D protein structure estimation, for example, are still very difficult problems, because none of the coordinates are even approximately independent of the others.

So you're back to a standard "minimization is really difficult" even in ultra-high dimensional spaces.