Formule
$Z_1 = W_1 X + b_1$, $A_1 = \sigma(Z_1)$, $Z_2 = W_2 A_1 + b_2$, $\hat{Y} = \sigma(Z_2)$ ; shapes : $X \in \mathbb{R}^{n_0 \times m}$, $W_\ell \in \mathbb{R}^{n_\ell \times n_{\ell-1}}$.
$Z_1 = W_1 X + b_1$, $A_1 = \sigma(Z_1)$, $Z_2 = W_2 A_1 + b_2$, $\hat{Y} = \sigma(Z_2)$ ; shapes : $X \in \mathbb{R}^{n_0 \times m}$, $W_\ell \in \mathbb{R}^{n_\ell \times n_{\ell-1}}$.