[futurebasic] Re: [FB] Random Number Generators

Message: < previous - next > : Reply : Subscribe : Cleanse
Home   : October 1999 : Group Archive : Group : All Groups

From: Robert Cogburn <rcogburn@...>
Date: Tue, 12 Oct 1999 00:53:45 -0600
>Many years ago, myself and a couple of other geeks thought of a way a
>create random number generator -- which was to use a radioactive counter.
>The number of "hits" per second was considerable -- in the ten of
>thousands. But then we would take the least significant number (to whatever
>number of digits you wanted) and we believed that number was truly random.
>No one to my knowledge has ever proved us wrong or right. What do you
>think? Is there some underlying periodic that would invalidate our solution?
>I can't help but think that any number generated by a computer could not be
>truly random unless it derived its randomness from nature.

This should be a very good way to generate a sequence with random
properties. The main problem with this for use in computations requiring
a lot of random numbers is that it is slow - typical computer generated
random numbers are produced at one or more per microsecond at G3 speeds.
Even Marsaglia acknowledges that the best way to produce high quality
results is by frequent seeding from a physical source, combined with a
high quality computer RNG. How to make this work well in practice is not
so clear. Still if you have the time a physical generator such as you
describe may be unbeatable. Of course you have the radiation to contend
with Š

Things like coins, dice, roulette wheels and lottery devices usually have
mechanical biases and problems with doing a sequence of trials without
other factors affecting results. Referring to Marsaglia someone said that
random numbers are worth their weight in gold but he is giving them away
(or something like that). The fact is they have lots of uses but it is
very difficult to produce them, especially in long sequences, and have
them stand up to scrutiny for randomness.

>We already incorporated Robert's suggestions in the compiler. You will see
>it in the next release.
>-STAZ  ~)~a

Whoops, what suggestions? Can't wait to see thisŠ

Of course, like belief in other things metaphysical, there are I think
more than a few already convinced of the compiler's random nature. Oh
wellŠ :)))