2

I am investigating a 2d Ising Spin Lattice.

I have been able to generate a Monte Carlo app that gives me the changing spin matrix through my iterations - like the many examples on the web.

However, I am trying to predict an independent variable (atmospheric pressure). I believe it is possible to measure the pseudo "Magnetism" and "Energy" of the timeseries using the Ising approach - but I am not sure how to do that. I assume it entails flipping the "spin" based on the timeseries changes, but I am not really sure weather that is the right approach.

If anyone could give me some advice that would be great.

EDIT 1

I have a series of pressure readings:

 Xt :1..N

Instead of using Metropolis flip where

deltaE = 2 * J * Si * sumOf(Sneighbours)
if (deltaE < 0 || exp(-deltaE/TEMP) > rand())
       Si+1 = -Si

I need to set 

Si+1 = someFunctionOf(Xt)

But I don't know what "someFunctionOf" should be.

If I did I could update the Lattice from Xt, and workout a "live" Magnetism and I am hoping that might tell something about the current phase of pressure changes.

ManInMoon
  • 121
  • 2
  • 1
    This sounds interesting, but I can't quite fill in the details myself. Could you make your analogy (and your question) a bit more precise? – user1504 Jun 11 '13 at 20:43
  • 1
    I am assuming that there is some map between your problem and an Ising model. However, if you are interested in time dependence, the answer will depend on the dynamics of your particular problem. There are many possible dynamics of an Ising model, all of which lead to the same equilibrium probability distribution of spins, p(S). The two "standard" models are Glauber dynamics (spins flip with some rate) and Kawasaki dynamics (spins exchange with some rate) - google them. However, if you are modeling an atmospheric problem, the dynamics will probably not resemble either of those. – AJK Jun 11 '13 at 23:17

0 Answers0