In hard svm all the points are at least 1 unit away from svm hyperplane. However, in soft svm we allow for slack of ksi_i for x_i. So x_i would/should be (at least) 1 - ksi_i units away from svm hyperplane.
@Yasser Othman Ideally the points should be at least distance 1 unit from svm hyperplane. In soft svm hyperplane this condition is relaxed by amount ksi_i units for x_i. So point x_i should be at least 1 - ksi_i distance from svm hyperplane. Thats why we are taking difference.
@@pankajkporwal So the points xj and xk are correctly classified, but they can not be called as support vectors? Only points on the margin that exhibit a distance of 1 unit from the hyperplane are called as support vectors?
Nice Explanation Sir. Great Video.
Excellent Video, thank you!
Very helpful explanation. Solved my query
amazingly useful! Thanks!
You are welcome!!! Please like, subscribe, and share!!!
@3:26 why did you subtract the slack variable from 1 ?
In hard svm all the points are at least 1 unit away from svm hyperplane. However, in soft svm we allow for slack of ksi_i for x_i. So x_i would/should be (at least) 1 - ksi_i units away from svm hyperplane.
@@pankajkporwal thanks but why not multiply or add instead of subtracted ?
@Yasser Othman Ideally the points should be at least distance 1 unit from svm hyperplane. In soft svm hyperplane this condition is relaxed by amount ksi_i units for x_i. So point x_i should be at least 1 - ksi_i distance from svm hyperplane. Thats why we are taking difference.
@@pankajkporwal So the points xj and xk are correctly classified, but they can not be called as support vectors? Only points on the margin that exhibit a distance of 1 unit from the hyperplane are called as support vectors?
For soft margin classifier, are the dots within the margin called support vectors? Or only the dots right at the margin are called support vectors?
Yes, only the point right at the margin are called support vectors.