ERRATA: The dual problem of the SVM objective says minimize, where it should say _maximize_.

We finish up the lecture by applying what we learned in the previous video to our SVM objective, and rewriting it to a form that contains only the KKT multipliers as parameters. This allows us to use the kernel trick, a powerful idea, that lets us train in high dimensional feature spaces, without explicitly computing lots of features.

slides: [ Ссылка ]
lecturer: Peter Bloem

свернуть