I am unable to get What is C here also the formula of Loss.plz make me clear of it
SVM -Handling Outlier
SVM loss function/objective i.e. Loss = ||w||^2, this does not allow any sort of leniency, for outliers and hence the hyper-plane received as the result will be more sensitive to outliers, and tries to classify them properly no matter how wrong are they.
Later on this loss function was changed to allow some outliers, i.e. Loss = ||w||^2 + c * sigma of max(0,1-ti) where ti = yi * ( w^T * xi + b). Now here c is the weight given to second half of the loss objective, meaning while minimizing our loss objective, how much influence it shares in the original loss function.
here we are now allowing some outlier that’s why we are increasing our loss function but how to know that we have to increase by c*(summationEi from 1 to m). Still didn’t understand Loss function.
You got the first half right, that to allow some outlier we are increasing our loss function to new one. Secondly we added this term c * (summation…), as it is suggested in “PEGASOS algorithm for svm”. This means that it is derived by mathematicians/researchers. Don’t think that you need to always derive new loss functions yourselves(unless you go in research area). Here we are just implementing the Pegasos algorithm and nothing more.
Also our task is to find the value of c for which our classifier performs better. C is a hyperparameter, and needs to be hypertuned with hit and trial.
ok got it .thanks…
I hope I’ve cleared your doubt. I ask you to please rate your experience here
Your feedback is very important. It helps us improve our platform and hence provide you
the learning experience you deserve.
On the off chance, you still have some questions or not find the answers satisfactory, you may reopen
the doubt.