Part $($d$-1)$

The difficulty in finding a good step size for convergence in gradient descent can often be attributed to the scale of the data. If the data is not standardized, variables with larger scales can dominate the gradient, making it hard for the algorithm to converge.

Recommended Step:

Scale

$$

X and

$$

y by the mean and standard deviation of each variable.

This step will standardize the variables, ensuring that each variable contributes equally to the gradient descent process.