$2.[5$ pts$]$ When a problem is non$-$linear separable, the Perceptron Training Rule never stops, while the

Delta Rule converges to a local optimum and stops. Explain the reason in detail using the update rule of

each method.