Observing the update step of SGD$,$

$larr-gra{d}_{}[Los{s}_h(y^{(i)}(*x^{(i)}+{}_0))+\frac{}{2}|||{|}^2]$

Which of the following is true?

As in perceptron, $$ is not updated when there is no mistake

Differently from perceptron, $$ is updated even when there is no mistake