Question 1 (50 points) AdaBoost. In AdaBoost, we showed that we shall choose t=21log(t1t) to minimize the upper bound of the training error. (a) If a function is a convex function, it has a global minimum. If a function's second derivative is always positive, it is a convex function. If we define the following function f()=e(1)+e, where [0,0.5). Show that f is a convex function. (b) Prove that the choice of t=21log(t1t) can minimize the training error's upper bound. Question 1 (50 points) AdaBoost. In AdaBoost, we showed that we shall choose t=21log(t1t) to minimize the upper bound of the training error. (a) If a function is a convex function, it has a global minimum. If a function's second derivative is always positive, it is a convex function. If we define the following function f()=e(1)+e, where [0,0.5). Show that f is a convex function. (b) Prove that the choice of t=21log(t1t) can minimize the training error's upper bound