ann_d_sum_of_sqr — derivative of sum-of-squares error
err_d Matrix containing the sum-of-squares error derivative, one per column, each column have a correspondent in y and t.
y Matrix containing the actual network outputs, one per column, each column have a correspondent in t and err_d.
t Matrix containing the targets, one per column, each column have a correspondent in y and err_d.
This function calculates the derivative of sum-of-squares error given y and t.
Any other, user defined, error function should have the same input and output format for variables.
Note: this function is extremly simple, it just calculates y-t.
The only reason to provide it is to allow the option to change the default error function if so is desired.