[SciPy-user] Maximum entropy distribution for Ising model - setup?

Martin Spacek scipy at mspacek.mm.st
Fri Oct 27 03:51:36 CDT 2006


Tim Leslie wrote:

> I could be wrong here, but doesn't the Ising model only consider
> nearest neighbour interactions? Would changing your setup to only
> consider these interactions make things work?

Yeah, that's what I've read as well, so I'm a bit confused over why the 
paper I mentioned calls this an Ising model. What we're doing is 
considering interactions among all pairs, although most of those are 
very weak. Anyway, what I need is to run maxentropy on this system:

P(s1, s2, ..., sN) = 1 / Z * exp( sum(hi*si) + 0.5*sum(Jij*si*sj) )
                                    i               i!=j

regardless of whether it's Ising or not. I think I can see how to do it 
in scipy.maxentropy if it were simply:

P(s1, s2, ..., sN) = 1 / Z * exp( sum(hi*si) )
                                    i

But getting it to work with the pairwise terms is just guesswork on my 
part...

Martin


More information about the SciPy-user mailing list