Re: Query: Spatially uncorrelated reproducible random values

[Sven N. Thommesen]

At 11:50 PM 8/2/2000 +0100, you wrote:
> Thanks, Paul.  That's certainly the kind of thing that I have in mind: 
> coordinates in, random value out.  However, I am not sure that 
> conventional random number generators generate values that are 
> uncorrelated with their seeds.  The usual tests for randomness are tests 
> of the non-correlation of successive values generated using the _same_ 
> seed.  Do you know that there is no correlation between a seed and the 
> (next) generated value, or is this just an assumption you've made?
>
> Nigel

Well, as the guy who put the "random" in Swarm, I can say that there is *no 
promise* that first values are uncorrelated with seeds. This might be the 
case with specific generators, but not by (my) design.

Also keep in mind that for a generator with a large state space (e.g. the 
default generator mt19937) a 32bit seed can only pick out a very small set 
of all possible starting states. [That's why I allow you to specify the 
state yourself, if you so choose.] So it's quite possible that two 
different seeds might produce the *same* 32-bit first variate.

As Nigel points out, the promise of random generators is usually that 
*successive values* will appear random, whatever the starting seed.

Cheers,
Sven Thommesen



                 ==================================
  Swarm-Modelling is for discussion of Simulation and Modelling techniques
  esp. using Swarm.  For list administration needs (esp. [un]subscribing),
  please send a message to <address@hidden> with "help" in the
  body of the message.
                 ==================================