I'm following the great course of stanford on ML, I was just wondering why the regularization term is different when using regression or classification. In regression we must add the term (lambda/m)*Theta And in classification we must add the term (lambda/2m)*Theta². I was so wondering about why these two term are different and why we're not using the same. Thanks

  • 2
    $\begingroup$ Please edit your question to say more about the context in which these symbols show up; a link to the Stanford web page(s) might help a lot. Is it possible that in the regression examples they are talking about something like LASSO with implicit variable selection, while with classification they are using something like ridge regression in which all variables are kept in the model but with penalized regression coefficients? $\endgroup$ – EdM May 23 at 16:01