We would like to use distanceweighted 2nearest neighbors to approximate the function f(x) = 8x - 10 - x2 given the data instances (x, f (13)): (1.0, -3.0), (3.0, 5.0), (5.0, 5.0), (7.0, 3.0). What is the value x = x0 at which the maximum error (Le f(x) - x is made in the approximation of f(x) in the region 3 S x 5 5 if we use distanceweighted 2-nearest neighbors? Would the error at 1'0 increase or decrease if we use 4-nearest neighbors with the given data? [5 MURI]