Is there any relation between k-means and the maximum-likelihood estimate in unsupervised learning? Any references would be appreciates!

  • $\begingroup$ What have you tried? Have you tried interpreting k-means as EM algorithm already? $\endgroup$ Nov 9, 2014 at 21:53
  • $\begingroup$ That was the idea.... I read about EM can be view as k-means generalization... But I can't find any proof. Can anyone help me? $\endgroup$
    – Kevin
    Nov 10, 2014 at 17:07
  • $\begingroup$ It's actually rather straightforward... first, have a look at e.g. Mclust in R, which is the R implementation of EM, and understand the model parameter; in particular the E model. $\endgroup$ Nov 10, 2014 at 18:31
  • $\begingroup$ Could you be more specific? Thank you so much! $\endgroup$
    – Kevin
    Nov 10, 2014 at 21:20
  • 1
    $\begingroup$ No fun when you are in a hurry :D $\endgroup$
    – Kevin
    Nov 12, 2014 at 1:44

1 Answer 1


You can think as K-Means as EM Algorithm (With Hard Thresholding) for GMM.
Since EM is a method for Maximum Likelihood Estimation you have a connection.


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.