Take the 2-minute tour ×
Mathematics Stack Exchange is a question and answer site for people studying math at any level and professionals in related fields. It's 100% free, no registration required.

Let $A \in \mathbb{R}^{n\times n}$ be a positive definite matrix. I want to evaluate the following integral. (This is part of a normalizer for a probability distribution.) $$ \int_{||u|| = 1} \frac{1}{\sqrt{u^\top A u}} \mathrm{d}u$$ This is the inverse of $A$ norm integrated over a the unit hypersphere. By a change of variable, $x = A^{1/2} u$, we can rewrite it as, $$ |A|^{1/2} \int_{||A^{-1/2} x|| = 1} \frac{1}{\sqrt{x^\top x}} \mathrm{d}x $$ which is an integral on a hyper-ellipsoid. Any hints appreciated!

[EDIT] In 2D, wolframalpha tells me that it's a Complete Elliptic Integral of the First Kind.

share|improve this question
    
perhaps, switch to n-spherical polar coordinates? –  Valentin Jan 26 '13 at 22:46
    
@Valentin Thanks for the suggestion. In n-spherical coordinates, the norm becomes simple, but interaction with $A$ is not very friendly, or is it? –  Memming Jan 26 '13 at 23:21
add comment

2 Answers

The first thing I would do is to do an orthogonal transformation:

$A = Q \Lambda Q^{\top}$, with orthogonal $Q$ and diagonal $\Lambda$.
The Q then cancels out and the integral is then $\int_{||x||=1} \frac{1}{\sum_{i=1}^d \lambda_i x_i^2} dx$
Not sure yet if it is nicely integrable thought (should look at $d=2$ and $d=3$ cases to get it)

share|improve this answer
add comment

I'm afraid that there might not be a good closed form solution. $\sqrt{u^TAu}$ has lots of branch cuts on the imaginary axis of the $u$ plane, which would be quite difficult to deal with. By using 4.646 of Gradshteyn and Ryzhik, I got something similar to $\sim\text{const}\cdot\int_0^\infty\frac{dx}{\sqrt{\det{(\mathbf{1}+x^2 A})}}$, but I am not sure how useful this could be.

share|improve this answer
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

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