Let A be an m x n matrix and b = abs(Ax) for an unknown ground truth vector x. We want to recover x from the absolute values of the linear measurements, b = abs (Ax). The approach in the homework is to minimize the loss function f(z): ||abs(Az) - b|| 2. There is another way to solve the same problem. Note that knowing b = abs (Ax) is equivalent to knowing (b) = (Ax). Here (-)2 is the component-wise square operation. Now we consider following minimization problem with the loss function g: g(z) = || (Az) (b) || 2: - = m ((a,z)2 b), i=1 = where ai is i-th row of the matrix A and b, is the i-th entry of b. (a) [2pts.] Show that the x is also the solution. More precisely, g(-x) = 0. (b) [5pts.] Let h(a) = (a- ) for some real number a and positive number 3. Show that h is not a convex function. This implies that the loss function g is not convex in general.