Suppose that f: [0,1]R is a continuous function and let be a positive real number. Then there exists a neural network with a single hidden layer output ONN such that for any x in [0,1]: |f(x)ONN| , generalize the theorem to functions defined on the following sets: (20 points)

3. [0, 1][0,1]R

4. [0,1][0,1]R²

5. [0,1]ⁿR