Cholesky decomposition of positive semidefinite matrix

Dear experts,

for a Monte Carlo study I need a Cholesky decomposition of a correlation matrix that is not positive definite, but positive-semidefinite. As I found out, one way is adjusting the matrix, an other way adjusting the method of computing the cholesky decomp. Unfortunately all algorithms I know to adjust a matrix only produce semi-definite matrices. Do you have a tip for me?

Thanks,

Albert
LVL 1
Albert-GeorgAsked:
Who is Participating?
I wear a lot of hats...

"The solutions and answers provided on Experts Exchange have been extremely helpful to me over the last few years. I wear a lot of hats - Developer, Database Administrator, Help Desk, etc., so I know a lot of things but not a lot about one thing. Experts Exchange gives me answers from people who do know a lot about one thing, in a easy to use platform." -Todd S.

GwynforWebCommented:
This a nasty problem to be dealing with. Cholesky decompositions of semi definite matrices are not unique, although do exist, and standard algorithms do not work.

One way of getting a solution is to multiple the diagonal elements by 1 + k where k>0 is really small, this will result in a +ve definite matrix.  Solving by standard means, for different values of small k>0, you can extrapolate the matrix values back to k=0. This will give one of the solutions, assuming convergence as k->0. You can then multiply the resulting matrices to see how well it corresponds to the original for error analysis.
0
Albert-GeorgAuthor Commented:
Dear GwynforWeb,

this sounds reasonable. I have read there is an alternative method to Cholesky decomposition which uses eigenvalues. Can you recommend this method?

Sincerely,

Albert
0
GwynforWebCommented:
Albert - I am not familiar with the eigenvalue technique. I don't envy you working with semi definite matrices , there are no really definitive algorithms or code around for dealing with them. I'd use what suits you best and is implementable in reasonable time. I'd play around with them and see what works, using the simplest first. The 'hack' I presented might well give sufficient accuracy for a single k = 10e-10 for instance.
0

Experts Exchange Solution brought to you by

Your issues matter to us.

Facing a tech roadblock? Get the help and guidance you need from experienced professionals who care. Ask your question anytime, anywhere, with no hassle.

Start your 7-day free trial
Albert-GeorgAuthor Commented:
Thank you very much. I will try out all possibilities and see what works.

Sincerely,

Albert
0
It's more than this solution.Get answers and train to solve all your tech problems - anytime, anywhere.Try it for free Edge Out The Competitionfor your dream job with proven skills and certifications.Get started today Stand Outas the employee with proven skills.Start learning today for free Move Your Career Forwardwith certification training in the latest technologies.Start your trial today
Math / Science

From novice to tech pro — start learning today.