[chuck-users] random seed
Lars Ullrich
mail at larsullrich.de
Sat Sep 11 09:31:17 EDT 2010
Hi Kas,
I was looking for a message to turn off the seed. Something like "set
seed to 0" and get non repeatable structures. (like maxmsp object
lp.tata from peter castine )
Anyway, I have found a method to do this with some functions.
Thanks a lot for your patience.
Lars
Am 11.09.2010 um 08:19 schrieb Kassen:
> Hey, Lars,
>
> Let me address your points in reverse order;
>
> > So the only way is to delete the seed and restart the sketch?
>
> Short answer; "no". Long answer; a PRNG (pseudo random number
> generator) typically consists of a number and a function. The
> function is set loose on the number and the outcome becomes the new
> number and is returned as the random value. Sometimes only part of
> the number, like the least significant bits are returned.
>
> It follows that these functions must make sure this series of
> numbers takes a long time before they repeat, that the "seed" simply
> determines at what part of the cycle we start and that talking about
> "deleting the seed" doesn't actually make a lot of sense.
>
>
>
> That's very sad! A seed is perfect to get a repeating pattern, but
> sometimes it's nice to turn off the seed to get non repeatable values.
>
>
>
> I'm not sure whether this is "sad"... it's simply a fact of life in
> computer science. The numbers are never really random and *if you
> knew the seed* they are easy to repeat. They just look random and
> for a decent PRNG it should be very hard to figure out what the seed
> was. This might be so hard that we'd all be gone by the time some
> process determines it.
>
> So; in practice this difference between repeatable numbers and truly
> random ones shouldn't matter too much.
>
> On the other side; it may very well be that something went wrong in
> the case of ChucK?
>
> Yours,
> Kas.
>
>
>
> _______________________________________________
> chuck-users mailing list
> chuck-users at lists.cs.princeton.edu
> https://lists.cs.princeton.edu/mailman/listinfo/chuck-users
More information about the chuck-users
mailing list