0
$\begingroup$

I am aware this could be a dumb question but I got myself stuck in some complicated calculus that I have greatly simplified to focus on my lack of understanding.

My goal is to compute the derivative of the following composition : \begin{align} (f\circ g)'(0) \end{align} Which is the composition evaluated at 0. $f: \mathbb{R^n}\rightarrow \mathbb{R}$ and $g: \mathbb{R}\rightarrow \mathbb{R^n}$ so the total function $f\circ g: \mathbb{R}\rightarrow \mathbb{R}$

Intuitively using the chain rule : \begin{align} (f\circ g)' = (f'\circ g)g' \end{align}

So in my particular case since $f$ is from $\mathbb{R^n}$ to derive it I need to take the sum of the partials multiplied by the derivative of the corresponding entry. For g I just need to derive each entry. The problem is that the derivative of g lead to n values that I can't evaluate on 0.

Can someone provide me a way to solve this ?

$\endgroup$
3
$\begingroup$

$g'$ has $n$ components, and $f$ has $n$ partial derivatives. You need to multiply each component of $g'$ with the corresponding partial derivative of $f$ (evaluated at $g(0)$) and then take the sum.

In general, the derivative of a function $\mathbb R^m\to\mathbb R^n$ is an $m\times n$ matrix, and the chain rule is a matrix multiplication.

$\endgroup$
  • $\begingroup$ In particular, $\mathrm df$ is a row vector and $\mathrm dg$ is a column vector here. $\endgroup$ – amd Nov 27 '15 at 22:33
  • $\begingroup$ I understand thank you ! So this has nothing to do with the total derivative of $f$ then ? On the wikipedia page of the chain rule, they say that the generalization in higher dimension involve the total derivative. $\endgroup$ – user149705 Nov 27 '15 at 22:41
  • $\begingroup$ As far as I understand total derivative it's a special case of the multidimensional chain rule. $\endgroup$ – Justpassingby Nov 28 '15 at 9:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.