Hi Igor,
The email was from me. I was trying to keep it high level enough for a bed time
story.
The purpose of the super high level description
to get "really random Q" take a "random" A and use DLARFP
was to avoid the nightmares if at all possible ;).
Cheers,
Julien.
On Nov 9, 2011, at 12:45 PM, KIA wrote:
On 09.11.2011 19:20, Langou, Julien wrote:
...........
On Nov 8, 2011, at 9:14 AM, Mark Gates wrote:
...........
DLARFP was requested by Alan Edleman, reason: this is how you should
construct "true" random orthogonal matrices:
take a random A, do a QR factorization, get the Q. That's a random
orthogonal matrix.
...........
Really random?! Consistent with the Haar measure on the group of
orthogonal matrices?!
(I supposed only one person could know the answer, that was Julie
Chislenko from Bank of New York, but she's not available any more.)
Besides, what is "random A"?
Mark, please, can You provide a link for a proof or a discussion of this
fact?
Sincerely Yours,
Igor Kondrashkin
