[[EM]]. ProbablisticParameterEstimation 방법

missing data를 estimation하는 general algorithm.

HiddenMarkovModel estimation 하는 BaumWelchAlgorithm 은 EM algorithm의 specilal case.
HMM의 경우 missing data는 unknown state.

statistical model 에서

parameter θ,
oberseved quantity x,
missing data y,

to find the model maximises the log likelihood
{{{
log P(x|θ) = log Sum(y)P(x,y|θ).
}}}

----
see also NoSmoke:ExpectationMaximization