In order to minimize the error, which method helps to find the derivative of the error with respect to each weight and subtracting this value from the weight value?