3
$\begingroup$

I saw in a paper (https://www.princeton.edu/~wbialek/rome/refs/bialek+ruyter_05.pdf Eq.37) the following identity:

enter image description here

where the <.> operator refers to a population average.

No source or justification was given for this identity & they seemed to assume the reader should be familiar with it.

However, I have no idea where they got this or how to proove it....

Any help please?

Thanks!

$\endgroup$
1
  • 3
    $\begingroup$ Smells like an integration by parts argument, applied to what one gets from multiplying out the LHS. $\endgroup$
    – Yemon Choi
    Mar 24, 2015 at 17:21

2 Answers 2

4
$\begingroup$

indeed, integration by parts does the trick:

$$\int d\mathbf{x}\; \mathbf{x} \,f(\mathbf{x})\exp\left(-\tfrac{1}{2}\mathbf{x}\cdot \mathbf{C}^{-1}\cdot\mathbf{x}\right)=-\int d\mathbf{x}\; \,f(\mathbf{x})\,\mathbf{C}\cdot\frac{\partial}{\partial \mathbf{x}}\exp\left(-\tfrac{1}{2}\mathbf{x}\cdot \mathbf{C}^{-1}\cdot\mathbf{x}\right)$$ $$=\int d\mathbf{x}\; \,\exp\left(-\tfrac{1}{2}\mathbf{x}\cdot \mathbf{C}^{-1}\cdot\mathbf{x}\right)\mathbf{C}\cdot\frac{\partial}{\partial \mathbf{x}}\, f(\mathbf{x})$$ $$\Rightarrow\left\langle\mathbf{x}\,f(\mathbf{x})\right\rangle=\mathbf{C}\cdot\left\langle \frac{\partial}{\partial \mathbf{x}}f(\mathbf{x})\right\rangle$$ where I have used that $\mathbf{C}$ is a symmetric matrix

$\endgroup$
2
  • $\begingroup$ Thank you Carlo. One question: why does the 'uv' part of integration by parts vanish? (here $u*v=\mathbf{C}f(\mathbf{x})\exp\left(-\tfrac{1}{2}\mathbf{x}\cdot \mathbf{C}^{-1}\cdot\mathbf{x}\right)$) $\endgroup$ Mar 24, 2015 at 23:41
  • 1
    $\begingroup$ you have to evaluate this boundary term at infinity, where it vanishes exponentially fast (assuming f(x) does not blow up at infinity, but then the average is not defined) $\endgroup$ Mar 25, 2015 at 7:14
2
$\begingroup$

This is the famous Stein's identity and indeed integration by parts is how you prove it!

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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