### Vos commentaires

*There are no independent solutions but solutions to problems.*

Some answers to your questions, apparently, they are trivial and therefore I am afraid I do not understand exactly the problem; so please expose some code and error/warning messages from Matlab.

So certainly the odds of finding solutions will grow.

Of course, I have nothing against :)

I sent to you by email a piece of reference from Matlab's childhood

It is useless, of course, but it is a relic from another time ... :)

Prior, to make it working, I used:

[z(perm,:),t] = ordschur(z,t,'lhp'); Success = 1;

if I had a thousand dollars for each data when I'm wrong, that would be great :) But how I'm not Microsoft employee's, I do not receive the money.

From zpotrf.f

" ...

A is COMPLEX*16 array, dimension (LDA,N)

On entry, the Hermitian matrix A. If UPLO = 'U', the leading

N-by-N upper triangular part of A contains the upper

triangular part of the matrix A,

leading N-by-N lower triangular part of A contains the lower

triangular part of the matrix A,

... "

What do you think?

" ...

A is COMPLEX*16 array, dimension (LDA,N)

On entry, the Hermitian matrix A. If UPLO = 'U', the leading

N-by-N upper triangular part of A contains the upper

triangular part of the matrix A,

__. If UPLO = 'L', the__**and the strictly lower**

triangular part of A is not referencedtriangular part of A is not referenced

leading N-by-N lower triangular part of A contains the lower

triangular part of the matrix A,

*.***and the strictly upper**

triangular part of A is not referencedtriangular part of A is not referenced

... "

What do you think?

I guess I'm missing something ... what does the norm matrix with problems raised by you, right or wrong behavior of the Matlab function chol?

Motto: "No matter what you want, what you get counts"

__example matrix was__

YourYour

**not**Hermitian !

*(but it does not really matter for matlab*

*chol! :) ... it's the truth )*

But ( back to Matlab doc):

"

`R = chol(A)`

__produces an upper triangular matrix R from the diagonal and upper triangle of matrix A__, satisfying the equation R'*R=A.

__The chol function__c. If it is not, chol uses the (complex conjugate) transpose of the upper triangle as the lower triangle. Matrix A ... "

**ASSUMES**that A is (complex Hermitian) symmetri*(his matrix A)*" ...must be positive definite."

**M**

**atlab chol function NEVER look your ENTIRE matrix**

**but ONLY**

**the**

**upper**

**triangular part**(or lower if the second arg is 'lower')

**of your matrix.**

__.__

**From this triangular part of YOUR matrix, Matlab build a virtual OWN matrix AND THAT MATRIX IS ALWAYS hermitian**(or symmetric if not complex)**The R result is always the decomposition of**

*THAT*HERMITIAN matrix.... and, YES, correct result should be R=[], p=1 and (

__I checked__

*yesterday even**symbolic*__)__, if you read my post I wrote " T

__he matrix, of course, is not positive defined but the documentation say__s: ...".

More, I suppose, with chances to be fair, the implementation of chol is a package of algorithms that are, when appropriate, improved.

However, the interface (the public contract of chol function) was maintained.

What I say is:

1.

that the result is fair in light of promises made in the matlab documentation and

2.

**I never said that sample matrix should be**

**positive**.

1.

2. but

*(assert( ~isempty ( chol(A, ...))))***NOT implies***( A***is**positive defined )2. but

*([~, p] = chol(A, ...) with p > 0)***implies***( A***is not**positive defined )Service d'assistance aux clients par UserEcho

absolutely, yes!