0
$\begingroup$

We are applying EM to a problem where we have N items to classify (binary classification, 0 or 1 ) for which a correct classification exists but is unknown, and a set J of "judges" that performs the classification. For example, a classification could consist in saying if a given picture contains an orchid or not. Each judge votes on a subset K

We don't know how "good" (accurate) each judge is in voting but we assume they are at least 50% accurate or more, that is, their answer is correct more than 50% of the time.

What we want to get out of this is the probability of each item being classified as 0 or 1, and for this we adopt the expectation maximization algorithm, which seems to be the way this is done in the literature. However, we also would like to get the credible interval for this probability. Can you help us or point us to a reference for how to compute this? Thanks

$\endgroup$
  • 1
    $\begingroup$ I think you are going to have to give us more details about the design. It might be good while editing your question to also expand EM to expectation maximisation in the title. $\endgroup$ – mdewey Jun 3 '17 at 12:42
  • $\begingroup$ @mdewey good point, I have done so, hope it makes sense now $\endgroup$ – Fab Jun 3 '17 at 13:32

Your Answer

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

Browse other questions tagged or ask your own question.