2. Show that, for the AdaBoost algorithm, the training data error rate will appromate to 0, given the sufficient training. Suppose: The training data:{(x',y', u')....(x", y",u")}, where x is the input vector, y (1 or u ^ -1) is the class label, and u is the weight; The final classifier: H(x) = sign a,f,(x)) ; Training data error rate: + 8(H(x) + y), where d' is the indicator function. N 11