Adding an implementation of the EM Algorithm to Turing for fitting latent variable models or mixture models

[00krishna]

Hello, after talking to @trappmartin on Slack, he suggested I open an issue. The request is to add an implementation of the EM (Expectation-Maximization) algorithm to Turing. The EM Algorithm is used to estimate latent variable models, mixture models, and other statistical models.

The GaussianMixtures package does seem to have a nice implementation of the EM algorithm. Perhaps the Turing developers could just create a wrapper around that package to keep the interface to the EM algorithm consistent with the Turing API.

[yebai]

Probably better to create a new EM package that supports Turing.