This is extension of the EM algorithm which typically converges more slowly than EM in terms of the iterations but can be much faster in the whole computer time. The general idea of the algorithm is to replace M-step of each EM iteration with the sequence of S >1conditional or constrained maximization or the CM-steps, each of which maximizes the expected complete-data log-likelihood found in the previous E-step subject to constraints on parameter of interest, θ, where the collection of all the constraints is such that the maximization is over the full parameter space of θ. Because the CM maximizations are over the smaller dimensional spaces, many times they are simpler, faster and more reliable than corresponding full maximization known in the M-step of the EM algorithm.