a loss function tells how good our current classifier is
You tell your algorithm what kind of errors you care about and what kind of errors you trade off against
Multi-class SVM loss
–– j could be the number of classes our dataset have
-syi – the sco
