Properties of zero-mean Gaussians

The following property of the Gaussian probability density functions (pdfs) is often used in this paper, here we state it in a form of a theorem:

Theorem 1   Let x $ \in$ $ \mathbb {R}$m and p(x) zero-mean Gaussian pdf with covariance $ \Sigma$ = {$ \Sigma_{ij}^{}$} (i, j from 1 to m). If g : $ \mathbb {R}$m $ \rightarrow$ $ \mathbb {R}$ is a differentiable function not growing faster than a polynomial and with partial derivatives

$\displaystyle \partial_{j}^{}$g(x) = $\displaystyle {\frac{\partial}{\partial x_j}}$g(x) ,

then

$\displaystyle \int_{{\mathbb R}^m}^{}$dxp(x)  xig(x) = $\displaystyle \sum_{j=1}^{m}$$\displaystyle \Sigma_{ij}^{}$$\displaystyle \int_{{\mathbb R}^m}^{}$dxp(x)  $\displaystyle \partial_{j}^{}$g(x) . (185)

In the following we will assume definite integration over $ \mathbb {R}$m whenever the integral appears. Alternatively, using the vector notation, the above identity reads:

$\displaystyle \int$dxp(x)  xg(x) = $\displaystyle \Sigma$$\displaystyle \int$dxp(x)  $\displaystyle \nabla$g(x) (186)

For a general Gaussian pdf with mean $ \mu$ the above equation transforms to:

$\displaystyle \int$dxp(x)  xg(x) = $\displaystyle \mu$$\displaystyle \int$dxp(x)  g(x) + $\displaystyle \Sigma$$\displaystyle \int$dxp(x)  $\displaystyle \nabla$g(x) (187)

Proof. The proof uses the partial integration rule:

$\displaystyle \int$dxp(x)$\displaystyle \nabla$g(x) = - $\displaystyle \int$dxg(x)$\displaystyle \nabla$p(x)    

where we have used the fast decay of the Gaussian function to dismiss one of the terms. Using the derivative of a Gaussian pdf. we have:

$\displaystyle \int$dxp(x)$\displaystyle \nabla$g(x) = $\displaystyle \int$dx  g(x)$\displaystyle \Sigma^{-1}_{}$xp(x)    

Multiplying both sides with $ \Sigma$ leads to eq. (186), completing the proof. For the nonzero mean the deductions are also straightforward. $ \qedsymbol$