Now, the question is how we can obtain the optimal w such that is minimized. We can do so using gradient descent: