解答
$$
\begin{aligned}
h_{nj}^{(1)} &= \text{sigmoid}\left(\sum_{i=1}^mx_{ni}w_{ji}^{(1)} + b_j^{(1)}\right)\\
\vdots\\
h_{nj}^{(l)} &= \text{sigmoid}\left(\sum_{i}h_{ni}^{(l-1)}w_{ji}^{(l-1)} + b_j^{(l-1)}\right)\\
y_{nj} &= \text{softmax}_j\left(\right)
\end{aligned}
$$