[EM] Path Voting Algorithm

Paul Dumais paul at amc.ab.ca
Fri Jun 11 11:20:30 PDT 1999


Markus Schulze wrote:
> 
> Dear Paul,
> 
> you wrote (11 Jun 1999):
> > Does anyone have a path voting algorith that they could post here? I'm
> > trying to code a MatLab program that runs path voting and Dumais voting
> > algorithms. I've just about got the Dumais algorithm figured out. It
> > seems that building a pair wise matrix is difficult. Dumais doesn't need
> > to build a pair wise matrix so, I've done it a different way.
> >     The difficulty in building a pair wise matrix is that even if you have
> > a vector (of size C! - almost 40 million with just C=11 candidates)
> > which represents the votes how do you code using loops etc so that you
> > can add the right elements that produce a particular pair wise entry? It
> > might be simpler to have an input matrix where the elements are the
> > margins or the borda scores. The columns could represent the candidates
> > while the rows would represent each vote. I'm doing it this way for
> > Dumais voting (but not creating a pair wise matrix). To get the
> > pair-wise matrix, we would iterate over the C(C-1)/2 possible pairs,
> > compare and sum. The matrix could then be used compute the (many?!)
> > paths to each opponent for each candidate. OK, I'm getting lost now. How
> > do you write an algorith that computes all possible paths? Path voting
> > seems really hard to code. I hope I'm wrong.
> >     I'll work on it this weekend. If anyone has done this work already,
> > please e-mail me on this list or at pdumais at powersurfr.com.
> 
> What is wrong with that algorithm I have posted yesterday?

Woops! Sorry, it looks good so far. I'll try it; thanks. Do you have the
matrix creation algorithm as well?
Are these algorithms posted on the web? I'll post my Matlab programs on
this list when I'm done (if anyone's interested).

-- 
Paul Dumais



More information about the Election-Methods mailing list