Derive the perceptron training rule and gradient descent training rule for a single unit with output , where = ₀ + ₁₁ + ₁₁²+ + + _n² . What are the advantages of using gradient descent training rule for training neural networks over the perceptron training rule?