THe problem is that the L^0 norm was being used and it was implemented assuming that the sparse vector data structure could be queried to find the number of zeroes.
This is, of course, incorrect. The data structure can only give an upper bound on the number of non-zero elements. The other part of the problem was that L^0 was probably not the right normalization to use. On Thu, May 28, 2009 at 12:07 AM, Sean Owen <[email protected]> wrote: > (Sorry if I am misunderstanding the question or calculation.) > > I think your point is, in a sparse vector, shouldn't we ignore the > 'fake' zeroes we observe where vectors are missing an element? >
