It can be used to fill the missing data in a sample. 
It can be used as the basis of unsupervised learning of clusters. 
It can be used for the purpose of estimating the parameters of Hidden 
Markov Model (HMM). It can be used for discovering the values of latent
variables.