1
$\begingroup$

How can the following integral be evaluated using Mathematica,

$\int_{\beta} \frac{1}{\left(||\mathbf{y}-\mathbf{D\beta}||^2 + \gamma\right)^{\alpha}} \exp\left(-\frac{1}{2}\beta^T\Sigma^{-1}\beta\right) \mathrm{d}\beta$,

where $\mathbf{\beta}$ and $\mathbf{y}$ are vectors of sizes $M\times 1$ and $N\times1$ respectively, $\mathbf{D}$ and $\mathbf{\Sigma}$ are matrices of respective sizes that allow the combination with vectors, $\alpha$ and $\gamma$ are positive integers and $||\cdot||$ denote the $\ell_2$ norm ?

$\endgroup$

closed as off-topic by Daniel Lichtblau, MarcoB, happy fish, Wjx, ubpdqn Mar 27 '17 at 7:40

This question appears to be off-topic. The users who voted to close gave this specific reason:

  • "This question cannot be answered without additional information. Questions on problems in code must describe the specific problem and include valid code to reproduce it. Any data used for programming examples should be embedded in the question or code to generate the (fake) data must be included." – Daniel Lichtblau, MarcoB, Wjx
If this question can be reworded to fit the rules in the help center, please edit the question.

  • $\begingroup$ I dont know if Mathematica will be able to do this independently of the dimensions $M$ and $N$. You can begin to build up your integration with M = 4; b = Array[bc, M]; int = Integrate[Exp[-b.b], b] and taking a look at Assuming. $\endgroup$ – Mauricio Fernández Mar 21 '17 at 16:23
  • $\begingroup$ Did you try to perform the usual analytic tricks like diagonalizing $\Sigma^{-1}$, transforming the integral measure and so on? $\endgroup$ – user46676 Mar 21 '17 at 16:28
  • $\begingroup$ @MauricioLobos: Indeed, building up seems to be a better approach. $\endgroup$ – Arun M Mar 21 '17 at 16:37
  • $\begingroup$ @marmot: Yes, I am trying with the simplest case of $\Sigma$ being the identity matrix, but it seems taking quite long time, and not yet converging to a closed form. However, I am a tenderfoot in making Mathematica evaluate these kind of integrals. $\endgroup$ – Arun M Mar 21 '17 at 16:37
  • $\begingroup$ @ArunM I guess this is more a maths than a mathematica problem since you can cook down the integral quite a bit by diagonalizing $\Sigma^{-1}$, i.e. $\Sigma^{-1}=U^T\cdot S\cdot U$ where $S$ is diagonal and positive (such that the integral converges), redefining $\beta$, $\beta'=\sqrt{S}\cdot U\cdot\beta$, and then` diagonalizing' $D$ in the $\beta'$ basis. I know that these steps are described in most of the literature on path integrals. $\endgroup$ – user46676 Mar 21 '17 at 17:01