Learning from data with EM algorithm

The learning of parameters in the framework of maximum likelihood can be performed with the Expectation-Maximization (EM) algorithm [Wolfe, 1970,Dempster et  al., 1977]. The following command trains (or learns) a model with the EM algorithm.

./bmix_train --data my.data --model-in \
     tiny.model --model trained.model \
     --iterations 50