The ClusterExpansion class currently solves the inverse Ising problem for a given threshold on changes in entropy for additional clusters [see John Barton and Simona Cocco, J. of Stat. Mech. P03002 (2013)].
A more user-friendly version would automatically vary this threshold until errors on pairwise statistics decrease below the expected fluctuations based on the amount of data (or some other user-specified error threshold).
A start on this code currently lives on the iterate-cluster-expansion branch, but it has not been heavily tested nor well-integrated into the coniii style.
The ClusterExpansion class currently solves the inverse Ising problem for a given threshold on changes in entropy for additional clusters [see John Barton and Simona Cocco, J. of Stat. Mech. P03002 (2013)].
A more user-friendly version would automatically vary this threshold until errors on pairwise statistics decrease below the expected fluctuations based on the amount of data (or some other user-specified error threshold).
A start on this code currently lives on the
iterate-cluster-expansion
branch, but it has not been heavily tested nor well-integrated into the coniii style.