Currently derivatives of functions with scalar inputs and vector outputs are called gradients, and the control flow is mixed in with
vector -> scalar functions.
In my opinion, this is confusing and error-prone.
Additionally, for the former an in-place f! can make sense while for the latter it does not.
Currently derivatives of functions with scalar inputs and vector outputs are called gradients, and the control flow is mixed in with vector -> scalar functions. In my opinion, this is confusing and error-prone.
Additionally, for the former an in-place
f!
can make sense while for the latter it does not.