Re: [infomap-nlp-users] Problems getting INFOMAP going

[Scott C. <ced...@gm...>]

Hi Sverker,

     Yes, I believe that the vectors *are* normalized post-SVD.  I
don't know offhand where this happens; I'm CC'ing the
infomap-nlp-users list in the hope that someone else does.  Otherwise
I'll try to figure this out when I get a chance... please prod me if
you haven't heard anything in a couple of days.


                             Scott

On 3/25/07, Sverker Sikstr=F6m <sve...@lu...> wrote:
>  Hi Scott!
>
>  I  think we have been in communication before about INFOMAP. I now have =
a
> the following specific question related to the output of INFOMAP. The wor=
d
> vectors that you get out of associate all seems to be normalized so the
> length of them are 1. However, I think that the SVD a algorithm (that i
> think INFOMAP is built on) do not neccessarily produce outputs where the
> vector lenght are 1, so I am inferring that somewhere in the program the
> vectors are normalized to one?! (please correct me if I am wrong!). I wou=
ld
> like to have access to length of the word vectors before normalization. D=
o
> you know how to do that? I looked into the associate c-code I could not f=
ind
> the normalization there so perhaps it is hidden somewhere in the
> 'infomap-build' command? I would appreciate an answer,
>
>  Sincerely, Sverker