That's great, it's clear now. Thank you so much for your help, Eibe.
Cheers,
Valerio
On 22 Dec 2017 5:08 p.m., "Eibe Frank" <eibe.frank(a)waikato.ac.nz> wrote:
Those two are different.
EM measures the joint log-likelihood. If x is an instance, and p(x) is the
estimated density for the instance, then log(p(x)) is what's used to
calculate the joint log-likelihood.
In contrast, the conditional log-likelihood is based on the conditional
probability of the observed class value y given the instance x, so it is
based on log(p(y|x)).
Cheers,
Eibe
On Fri, Dec 22, 2017 at 7:57 PM, Valerio jus <valeriojus(a)gmail.com> wrote:
Thanks a lot, Eibe.
It's the scheme's negative conditional loglikelihood (aka log loss),
computed from the class probabilities of the test
instances using base-2
logarithm.
Kindly I have this observation and need your valuable opinion about it. I
observed the appearance of the term "Log likelihood" in a cluster method
like EM. Therefore, how similiar this "Log likelihood" to the one you
mentioned (ie., loglikelihood or log loss). Are the two terms refer to one
thing?
Valerio
The negative conditional loglikelihood on the training data is what
logistic regression minimizes. You can view logistic regression as
empirical risk minimization if you measure risk using log loss, and you
consider logistic regression models as your set of hyotheses.
Cheers,
Eibe
On Thu, Dec 21, 2017 at 10:00 PM, Valerio jus <valeriojus(a)gmail.com>
wrote:
Dear Eibe,
Thanks for the detaild reply.
1- However, in your opinion, can the "Class complexity | scheme " of
entropy help us in finding/computing the "Empirical risk minimization"?
2- How can one interpret the "Class complexity | scheme" ?(how to use
it for the sake of scheme evaluation?)
Thanks.
Valerio
_______________________________________________
Wekalist mailing list
Send posts to: Wekalist(a)list.waikato.ac.nz
List info and subscription status:
https://list.waikato.ac.nz/mai
lman/listinfo/wekalist
List etiquette:
http://www.cs.waikato.ac.nz/~m
l/weka/mailinglist_etiquette.html
_______________________________________________
Wekalist mailing list
Send posts to: Wekalist(a)list.waikato.ac.nz
List info and subscription status:
https://list.waikato.ac.nz/mai
lman/listinfo/wekalist
List etiquette:
http://www.cs.waikato.ac.nz/~m
l/weka/mailinglist_etiquette.html
_______________________________________________
Wekalist mailing list
Send posts to: Wekalist(a)list.waikato.ac.nz
List info and subscription status:
https://list.waikato.ac.nz/mai
lman/listinfo/wekalist
List etiquette:
http://www.cs.waikato.ac.nz/~m
l/weka/mailinglist_etiquette.html
_______________________________________________
Wekalist mailing list
Send posts to: Wekalist(a)list.waikato.ac.nz
List info and subscription status:
https://list.waikato.ac.nz/
mailman/listinfo/wekalist
List etiquette:
http://www.cs.waikato.ac.nz/~
ml/weka/mailinglist_etiquette.html