Run gradient descent for $100$ steps with constant step size $=1/4.$ Compare two

different initialization: initialization $$A$$

w$1=2\mathrm{.}5,$ w$2=3\mathrm{.}56$

and initialization $$B$$

w$1=2\mathrm{.}5,$ w$2=3\mathrm{.}57.$

Answer the following questions:

$1.$ Which initialization led to better performance in terms of the training error?

$2.$ Is one of the converged final parameter $($nearly$)$ optimal? How do you know?