Refer to the following function and a one layer neural network. Activation function of hidden nodes is sigmoid and output node uses linear activation function.

We want to approximate the function in figure $($a$)$ using red bars $($e$.$g$.,$ A and B$).$

$$

$$

$($a$)($b$)$

$$

$1)[4$ pts$/$ea$]$ Now we want to implement the red bar $$A$$ using a neural network in figure $($b$).$ Show all the parameters of hidden nodes h$1$ & h$2($e$.$g$.,$ the weights, bias, slope of curve$),$ respectively. Show the exact shape of output of each hidden node.

$$

h$1$:

$$

h$2$:

$$

$2)[4$ pts$]$ Show all the parameters of output node. Explain in detail how the function is implemented using this network.

$$

$3)[4$ pts$]$ Show the output value of network given an input $2\mathrm{.}5.$

$($a$)$

$($b$)$