Symmetric Gaussian?

[keatonb]

Hello, thank you for this very useful package. It looks like it can do very nearly what I need for a project, but I am hoping that it may be possible to constrain the shape (but not scale) of a Gaussian component. Specifically, I'd like a Gaussian to be circular, with covariance proportional, but not equal, to [[1, 0],[0,1]]. Doesn't look like pygmmis can do that, but thought I'd ask since I saw in #1 that there was some built-in functionality that was just not exposed to the fit function. Cheers!

[pmelchior]

That's not trivial to implement in a general EM algorithm because it would alter the M step. However, one could try something like a proximal projection. That is, after a few iterations, replace the covariance with the nearest one (in the Euclidean sense) that is circular with some scalar size. Then run EM again for a few steps, or just one. Rinse and repeat.