C$3\_1$part$2$:

Try implementing a different loss function for synthetic regression data, such as the

absolute value loss $(y\_$hat $-$ d$21.$reshape$(y,y\_$hat.shape $)).$abs$().$sum$().$

Check what happens for regular data.

Check whether there is a difference in behavior if you actively perturb $|$

some entries, such as $y_5=10000,$ of $y.$

Can you think of a cheap solution for combining the best aspects of

squared loss and absolute value loss? Hint: how can you avoid really

large gradient values?

please use python