Expectation maximization iteratively produces an estimate of sufficient statistics for learnable parameters, then maximizes the parameters according to the estimate.
Methods for creating probabilistic factors associated with elements and their sufficient statistics.