In the EM algorithm, what is the goal of the 'M-step'?
To estimate the missing data given the current parameter estimates.
To maximize the expectation of the complete-data log-likelihood calculated in the E-step with respect to the parameters.
To calculate the probability of the observed data given the parameters.
To determine the number of iterations required for convergence.