Hi,
When learning a GaussianMixtureModel it should be possible to pass MaxIterations to the EM algorithm. As it is now the Iterations parameter is passed to EM instead of MaxIterations.
Since Iterations are private set and never changed in code it will always be 0. This leads to infinite loops when a problem does not converge.

I also notice a missing else statement in MultivariateNormalDistribution.Fit().
When computing unweighted covariance it will always default to full matrix even when options.Diagonal is set:

// Compute covariance matrix
if (options != null && options.Diagonal)
    cov = Matrix.Diagonal(Measures.Variance(observations, means));
cov = Measures.Covariance(observations, means).ToMatrix();

Best regards,
Frank