model estimation: expectation-maximization (EM) (Dempster et al. ’77)
- iterate between E-step and M-step until the likelihood of the data stops increasing.
- expectation step: assume the model (cluster parameters) to be correct and find the most likely distribution of the data with respect to the model!
- maximization step: assume the distribution to be correct and reevaluate the model parameters to maximize the likelihood of the data (cluster parameters) with respect to the model!