izmailovpavel/flowgmm

about gmm weight

akidway opened this issue · 1 comments

Hi,

Thank you for your great work.

In the code here, it appears that you compute gmm weights directly using a gradient-based method, which differs from the traditional EM method. Could you confirm if my understanding is correct?

Thank you!

Hey @akidway, thank you for checking out the code! I believe we explored both EM-style updates and direct optimization with SGD, and found that SGD works similarly or better.