How are these derivatives computed? It is part of this MOOC :
I can't understand the dimensionality of $\frac{\partial a^{<t>}}{\partial W_{a x}}$ , with $a^{<t>} \in R^{n_a}$ et $W_{a x} \in R^{n_a \times n_x}$, for me it would be a tensor of dimension 3 since we derive a vector with respect to a matrix.
But in the extract $\frac{\partial a^{<t>}}{\partial W_{a x}}=\left(1-\tanh \left(W_{a x} x^{(i)}+W_{a a} a^{<t-1>}+b\right)^{2}\right) x^{<t> T}$: the left member is a matrix (tensor with 2 dimensions).