Incomplete covariance matrices about pycave HOT 6 CLOSED

Hey @esrah-ap-niht. Unfortunately, full covariance matrices are currently not implemented. I'm in the process of refactoring the library though to make the interface even closer to sklearn's. I'll make sure to add it then :)

from pycave.

Hi @esrah-ap-niht ,
Are you using diagonal covariance?

from pycave.

@groadabike
No, I had gmm = GMM(num_components=20, num_features=test2.shape[1], covariance='spherical') specified. However, you are absolutely correct that the specification explains the shape of the matrix. I feel stupid I didn't catch that myself.

Thinking about this, the thread title should actually change to a feature request for a full covariance matrix option. Or if that's already implemented, knowing what text to use in the specification. I tried using "full" in the same style as Sklearn and received an error message.

from pycave.

Thank you @borchero, I'll look forward to it. I'm using your code for research and would like to cite your work. Do you have a reference you'd like me to use? Otherwise I can use a generic reference along the lines of:

from pycave.

Happy to see it being used in your research, @esrah-ap-niht ! :D unfortunately, there is no public work using the library, so the generic reference is definitely fine :) Thanks for asking!

from pycave.

@esrah-ap-niht full covariances matrices are finally implemented :)

from pycave.