Deriving the delta rule
	• Define the error as the squared residuals summed over all training cases: • Now differentiate to get error derivatives for weights • The batch delta rule changes the weights in proportion to their error derivatives summed over all training cases