Learning conditional PDF’s
	This doesn’t work as well in higher-dimensional spaces.  Why not?