Ising Models
An Ising model is an array of spins (atoms that can take states ) that are magnetically coupled to each other. If one spin is, say
$+1, is energetically favourable for its immediate neighbours to be in
the same state.
Let the state of an Ising model with spins be a
vector in which each component takes values and . If
two spins and are neighbours, we say .
The coupling between neighbouring spins is . We define
if and are neighbours, and otherwise. The energy
of a state is:
where is the applied field. The probability that the state is
is:
where , is the Boltzmannβs constant,
and:
The Ising model is also an example of a Markov Random Field (MRF). We
create a graph in the form of a 2D or 3D lattice, and connect
neighbouring variables. We can define the following clique potential:
Here is the coupling strength between nodes and . If
two nodes are not connected, . The weight matrix
is symmetric: . All edges have the same
strength , where .
The Ising model is analogous to the Gaussian graphical models. First,
assuming , we can write the unnormalized log
probability of an Ising model as follows:
where , and is the
bias term, corresponding to the local fields. If we define:
we can rewrite this in a form that looks similar to a Gaussian: