correlation between generators with default state

[David Bateman]

Brian Gough wrote:
> Currently, if no state argument is supplied all the MT19937-based
> generators are initialised with the same state, so they give
> correlated samples.
>
> To avoid the correlations I'd suggest initialising them independently
> (patch below).
>
> The attached picture shows the first values of randn() and rande()
> starting Octave 1000 times.  
>
>   
>
Uggghhhh..... I'm all for applying this patch as the correlation is 
visually very obvious even with this limited sample.

D.


-- 
David Bateman                                David.Bateman at motorola.com
Motorola Labs - Paris                        +33 1 69 35 48 04 (Ph) 
Parc Les Algorithmes, Commune de St Aubin    +33 6 72 01 06 33 (Mob) 
91193 Gif-Sur-Yvette FRANCE                  +33 1 69 35 77 01 (Fax) 

The information contained in this communication has been classified as: 

[x] General Business Information 
[ ] Motorola Internal Use Only 
[ ] Motorola Confidential Proprietary