I have been studying SVM lately, following Andrew Ng's CS229 lecture notes. I can understand most of the notes. But for the case where the KKT condition is satisfied at alpha = C, I am not sure what that means.

I know that for alpha = 0, the KKT condition is satisfied with inequality constraint, namely, the point lies above the decision function yi * (w' * xi + b) = 1. And for 0 < alpha < C, the KKT condition is satisfied with equality constraint, namely, the point is a support vector. Now as for the case where alpha = C, I wonder what does this mean, does it mean that the point violate the constraint and that the penalty is C * slack_variable?


1 Answer 1


$C$ is the maximum and $\alpha$ can be and indicates the Support Vector is inside the margin.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.