In the EM algorithm, what is the purpose of the 'M-step'?
To maximize the expected log-likelihood found in the E-step to update the parameter estimates.
To calculate the posterior probability of missing data.
To initialize the parameters of the model.
To check for convergence of the likelihood.