Suppose that a training set contains only a single example, repeated 100 times. In 80 of the 100 cases, the single output value is I; in the other 20, it is 0. What will a back- propagation network predict for this example, assuming that it has been trained and reaches a global optimum?