Under the plots comparing “unnormalized”, “X - mu”, and “Z score normalized”, the description for the rightmost plot says
Right: The second step divides by the variance. This leaves both features centered at zero with a similar scale.
However, excerpting from the previous code block
sigma = np.std(X_train,axis=0)
...
X_norm = (X_train - mu)/sigma
...
ax[2].scatter(X_norm[:,0], X_norm[:,3])
So it looks to me like we’re dividing by the standard deviation (sigma), not the variance (sigma^2). The previous discussion talks of dividing by the standard deviation. Should the description of the the plots read the following way instead?
Right: The second step divides by the standard deviation. This leaves both features centered at zero with a similar scale.
Or am I missing something?
Thanks for any clarification.
