3
$\begingroup$

I fit EM algorithm to mixture data. Then, I would like to implement AIC for this mixture data. I really could not find a clear mathematical formula to implement AIC for mixture data when we use EM algorithm. So, based on my reading, I understand that:

$$AIC = -2\log(\hat{\theta}) + 2K$$

where K is the number of model parameters.

$$\log(\hat{\theta}) = \sum (\tau_1(\log(\pi_1)+\log(c_1(\theta))) + (\tau_2(\log(\pi_2)+\log(c_2(\theta)))$$

where $c_1$ and $c_2$ are the first and second mixture components (densities) respectively.

Is that correct? Or do I did something wrong?

$\endgroup$
1
$\begingroup$

The AIC in this case can be obtained from the expected log likelihood.

See:

Use of the AIC with the EM algorithm: A demonstration of a probability model selection technique.

https://www.osti.gov/scitech/servlets/purl/86954

$\endgroup$
  • 1
    $\begingroup$ Could you expand on this answer here? It would help future readers, especially if the link goes dead. $\endgroup$ – mkt Nov 8 '17 at 4:00
  • $\begingroup$ Could you please gives further explanation. $\endgroup$ – Silver_80 Nov 8 '17 at 7:53

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.