= Solution
With $f=\mathbf1_{\{\mathrm{font}=\mathrm{serif}\}}$ and $d=\mathbf1_{\{\mathrm{display}=\mathrm{popup}\}}$, the model-matrix input is
$$
x=(f,d,fd)^T\in\mathbb R^3.
$$
The <feedforward neural network> has three inputs, a fully connected layer of two <rectified linear unit>[ReLU] units, and a fully connected two-class <softmax function>[softmax] output. Algebraically,
$$
h=\operatorname{ReLU}(Wx+b),
\qquad
z=Vh+c,
\qquad
p_k=\frac{e^{z_k}}{e^{z_0}+e^{z_1}},
$$
where $W\in\mathbb R^{2\times3}$, $b\in\mathbb R^2$, $V\in\mathbb R^{2\times2}$, and $c\in\mathbb R^2$. The coding is $0$ for no click and $1$ for a click. There are
$$
2(3)+2+2(2)+2=14
$$
parameters.
Back to article page
