Scaling

Three different techniques for feature scaling:

• Feature scaling, essentially dividing each positive feature by its maximum value, or more generally, rescale each feature by both its minimum and maximum values using (x-min)/(max-min). Both ways normalizes features to the range of -1 and 1, where the former method works for positive features which is simple and serves well for the lecture's example, and the latter method works for any features.
• Mean normalization: 𝑥𝑖:=𝑥𝑖−𝜇𝑖𝑚𝑎𝑥−𝑚𝑖𝑛
• Z-score normalization:

To implement z-score normalization, adjust your input values as shown in this formula:

(𝑖)𝑗=𝑥(𝑖)𝑗−𝜇𝑗𝜎𝑗where