Hi Dmitry,
At first glance, I'd say this is probably not a bug. GMM doesn't solve the
problem of "what exact Gaussian distributions generated my data". It solves
the problem of "what mixture of Gaussians minimizes this cost function".
I'd suspect that for the data you're using, the cost function a
To scikit-learn-general,
I fit the bimodal 1D distribution with the strong overlap of Gaussian
components using scikits.mixture.GMM. The scikits.mixture.GMM.fit gives
result which is inconsistent with parameters of input distribution.
The code below demonstrates the issue.
In case the two compon