1pt - True or False Multilayer feedforward network with at least one hidden layer are universal approximators (i.e. they can learn any given function with arbitrary accuracy). 1pt True or False - Using 3 or more hidden layers is necessary when the problem is very difficult or discontinuous. 1pt True or False There are some Boolean functions that can not be implemented with a one-hidden-layer feedforward network? 1pt True or False You need to use a linear activation function at the output unit, for regression problems. 1pt True or False A one hidden layer feedforward network using only linear activation functions is equivalent to a network with no hidden layers (i.e. there is no gain in using the hidden layer). iv) 2pt Give the formula and/or explain the gradient of a function f of 3 variables, x, y, and z.