Parameter Derivatives

$2$ points possible $($graded$)$

During SGD we are interested in relating the errors computed by back$-$propagation to the quantities of real interest: the partial derivatives of the loss with respect to our parameters. Here that is and $.$

What is $?$ Write in terms of the variables $,,,$ and if necessary.