Abstract
Computing marginal distributions of discrete or semi-discrete Markov RandomFields (MRF) is a fundamental, generally intractable, problem with a vastnumber of applications on virtually all fields of science. We present a newfamily of computational schemes to calculate approximately marginals ofdiscrete MRFs. This method shares some desirable properties with BeliefPropagation, in particular providing exact marginals on acyclic graphs; but atdifference with it, it includes some loop corrections, i.e. it takes intoaccount correlations coming from all cycles in the factor graph. It is alsosimilar to Adaptive TAP, but at difference with it, the consistency is not onthe first two moments of the distribution but rather on the value of itsdensity on a subset of values. Results on random connectivity and finitedimensional Ising and Edward-Anderson models show a significant improvementwith respect to the Bethe-Peierls (tree) approximation in all cases, and withrespect to Plaquette Cluster Variational Method approximation in many cases. Inparticular, for the critical inverse temperature $\beta_{c}$ of the homogeneoushypercubic lattice, the expansion of $\left(d\beta_{c}\right)^{-1}$ around$d=\infty$ of the proposed scheme is exact up to the $d^{-4}$ order, whereasthe two latter are exact only up to the $d^{-2}$ order.