I want to show that the updating in the Kernel perceptron $($with a$\_$i $=$ a$\_$i $+1)$ with a margin is equivilant to gradient descent of a single example and parameter of the hinge loss