Now you will implement the Pegasos algorithm. For more information, refer to the original paper at original paper.

The following pseudo-code describes the Pegasos update rule.

Pegasos update rule((),(),,,): if ()(())1 then update =(1)+()() else: update =(1)

The parameter is a decaying factor that will decrease over time. The parameter is a regularizing parameter.

In this problem, you will need to adapt this update rule to add a bias term (0) to the hypothesis, but take care not to penalize the magnitude of 0.

Pegasos Single Step Update

1 point possible (graded)

Next you will implement the single step update for the Pegasos algorithm. This function is very similar to the function that you implemented in Perceptron Single Step Update, except that it should utilize the Pegasos parameter update rules instead of those for perceptron. The function will also be passed a and value to use for updates.

Available Functions: You have access to the NumPy python library as np.