# can I fit a GMM over probability distributions?

We have been using a GMM to fit gaussians over a set of ~ 3M vectors.

Now the input are not vectors but probability distributions (coming from a topic model like LDA). Is it still mathematically sound to fit/train a GMM over such input?

• What do you mean by probability distributions? How are you representing it in the code without vectors? – Souradeep Nanda Jan 20 '17 at 16:32