SVM ng Courseware

1 .???? Margins margin

Functional margin

Consider the Sigmoid Function

If X> 0 is bigger and closer to 1, we have more confidence that the prediction result is 1. If x <0 is smaller, we are more likely to think that the prediction is 0,

If X is close to the origin, the prediction result is about 0.5 less credible.

Obviously, if we represent a positive sample, we have a lot of confidence to predict, but we are not very sure about it ????

?

SVM ng Courseware