We are trying to explore and solve a problem where two data clouds (dis-
tributions) in hyperspace need to be compared. We consider the need to
compute the /overlap/ between the two clouds, but the problem is hard be-
cause we deal with a high dimensional space (>100 dims) and a small number
n of sample points (100<~n<~1000).
Is there any known approach for doing this? Could anybody please point us
to a paper on the subject? So far we have computed the Euclidean and shuf-
fle distance between each point that forms one cloud with each point in
the other cloud.
Many thanks,
Roy
|
|