ConclusionΒΆ
From a subset of variables, we recovered successfully (i) the observed-to-observed, hidden-to-observed, observed-to-hidden, and hidden-to-hidden interactions; (ii) the configurations of hidden variables; (iii) the number of hidden variables.
The performance of FEM is significantly more accurate than other existing methods such as mean field approximations and Maximum Likelihood Estimation (MLE). Furthermore, FEM works approximately 100 faster than MLE.