There is a variant of boosting called gentleboost. How does gentle boosting differ from the better-known AdaBoost?
$\begingroup$
$\endgroup$
3
-
$\begingroup$ What other variants? It might be helpful to tighten up this question a little (more specificity). $\endgroup$– ShaneJul 20, 2010 at 16:03
-
$\begingroup$ Right; it is just one variant among others. $\endgroup$– user88Jul 20, 2010 at 17:11
-
$\begingroup$ OK, I'll ask for comparison to adaboost since that is perhaps the best known. $\endgroup$– Vebjorn LjosaJul 20, 2010 at 19:24
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
The second paper you cite seems to contain the answer to your question. To recap; mathematically, the main difference is in the shape of the loss function being used. Friedman, Hastie, and Tibshirani's loss function being easier to optimize at each iteration. –
