tag:blogger.com,1999:blog-1374054023917435198.post5690909187568915547..comments2024-03-25T21:39:34.690-07:00Comments on Musings about machine learning and other things: Discriminative vs. generative learning: which one is more efficient?Csaba Szepesvárihttp://www.blogger.com/profile/13790307935040509983noreply@blogger.comBlogger3125tag:blogger.com,1999:blog-1374054023917435198.post-32922895740254841672010-12-18T06:21:24.632-07:002010-12-18T06:21:24.632-07:00Some of the formulas you've mentioned througho...Some of the formulas you've mentioned throughout the article were a bit out of my league. I'm a newbie when it comes to logistic regression. Despite that, it was fascinating reading your post.moving companyhttp://www.moishes.comnoreply@blogger.comtag:blogger.com,1999:blog-1374054023917435198.post-69125963095130559102007-08-13T05:55:00.000-07:002007-08-13T05:55:00.000-07:00Thanks for the link -- the paper is indeed interes...Thanks for the link -- the paper is indeed interesting and relevant. I am not sure that the term discriminative training should be abolished though, as suggested at the end of the paper. In particular, I strongly disagree with the statement that "there is only one correct way to train [it]". However, the interpretation of discriminative training as a maximum likelihood method with a more flexible model is interesting. Rubinstein and Hastie also mention this.<BR/><BR/>And you were right: the log was missing. Now I corrected the post.Csaba Szepesvárihttps://www.blogger.com/profile/13790307935040509983noreply@blogger.comtag:blogger.com,1999:blog-1374054023917435198.post-49230777449217233392007-08-13T04:23:00.000-07:002007-08-13T04:23:00.000-07:00Interesting post!You could be interested by this p...Interesting post!<BR/>You could be interested by <A HREF="ftp://ftp.research.microsoft.com/pub/tr/TR-2005-144.pdf" REL="nofollow">this paper</A>.<BR/>PS: I think you forgot a "log" in the formula after "The meaning of this is that while naive Bayes maximizes the total joint log-likelihood"...<BR/><BR/>Best regards,<BR/><BR/>PierreAnonymousnoreply@blogger.com