Re: [Help-gsl] copying the symmetric part of a matrix necessary?

[Brian Gough]

At Mon, 14 Aug 2006 17:52:01 +0200,
Mstislav Elagin wrote:
> Later on, I need the Cholesky factorization of S. However, the
> procedure gsl_linalg_cholesky_decomp does not accept a parameter of
> type CBLAS_UPLO_t, so I assume it awaits a _full_ symmetric matrix,
> not only the upper or lower part. This looks strange to me, since the
> Cholesky decomposition is only defined for symmetric matrices and it
> would be logical to let the user supply a partly stored matrix
> obtained from another GSL procedure.
> 
> Do I actually need to copy elements from one part to the other in
> order to pass the matrix to gsl_linalg_cholesky_decomp? How do I do
> this with GSL short of explicit loops?

Although it is not documented, only the lower triangular part of the
matrix is used in the Cholesky decomposition, so it is safe to pass
that.  I will add a note to the manual about it.

-- 
best regards,

Brian Gough

Network Theory Ltd,
Publishing the GSL Manual - http://www.network-theory.co.uk/gsl/manual/