If I'm asked to differentiate a function that is simply of the form: $ f(tx_1, tx_2, ... tx_n) $ (i.e., the function is not "defined" to be particularly anything), how should I go about this? I want to differentiate with respect to $ t$.

Context: Let's say we have a homogeneous function of with degree $ k $.


$$ f( tx_1, tx_2, ... tx_n) = t^k f(x_1, x_2, ... x_n) $$

Differentiating on both sides with respect to $ t$, we get,

$$ x_1 f_1 (tx_1, tx_2, ... tx_n) + ... + x_n f_n (tx_1, tx_2, ... tx_n) = k \cdot t^{k-1} \cdot f(x_1, .. x_n) $$

  1. On the RHS, why we differnetiate with respect to to get $ k \cdot t^{k-1} $. Because the function on the RHS is a constant (i.e., it doesn't depend on t, we write it as it is)

  2. On the LHS, why do we have: $ x_n f_n(tx_1, tx_2,\dots, tx_n)$ for each term?

  • $\begingroup$ Shouldn't your RHS be $k \cdot t^{k-1} \cdot f(x_1, \dots, x_n)$? (without the $t$ inside) $\endgroup$ – BigbearZzz Jan 17 at 11:22
  • $\begingroup$ Yes, sorry. That was a mistake on my part. $\endgroup$ – WorldGov Jan 17 at 11:23
  • $\begingroup$ You need to assume that $k>1$. For $k>1$, you may fact out the variable $x_1$ first, then factor out the variable $x_2$ from the remainder and so on. Try to write down an example and you'll see it is simple. $\endgroup$ – James Jan 17 at 11:27

On the RHS, you just apply the usual rules of single variable calculus. The factor $t^k$ becomes $k t^{k-1}$ and $f(x_1,\dots,x_n)$ is constant with respect to $t$, so you can just keep it.

For the LHS you are looking for the chain rule for multivariable functions. If $f(x_1,\dots,x_n)$ is a function of $n$ variables and you have $n$ functions $z_1(t),\dots,z_n(t)$ each of one variable $t$, then

$$ \frac{\mathrm d}{\mathrm dt} f(z_1(t),\dots,z_n(t)) = \frac{\partial f}{\partial x_1}(z_1(t),\dots,z_n(t)) \cdot \frac{\mathrm dz_1}{\mathrm dt}(t) + \dots +\frac{\partial f}{\partial x_n}(z_1(t),\dots,z_n(t)) \cdot \frac{\mathrm dz_n}{\mathrm dt}(t) $$

In your example you have $z_i(t) = tx_i$ so that $\frac{\mathrm dz_i}{\mathrm dt}(t) = x_i$.


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.