random

Darren New dnew at san.rr.com
Fri Jun 1 19:22:33 CEST 2001


Alex Martelli wrote:
> it started with just the idea of defining randomness in
> terms of minimal amount of information (in bits) you need
> for prediction.  If YOUR definition of 'randomness' is
> something that needs INFINITE number of bits to predict,
> go ahead and have fun, but you're unlikely to get some
> constructive results out of it.

Actually, there's apparently two common "randomness" words that mean
different things. Chaitin is talking about finite-length random numbers.
These are "random" if you can't compress them, very basically. 

Others talking about random talk about random processes, which is
unbounded in output length. Random numbers are numbers from a random
process. A number isn't random in and of itself, but rather random if it
comes from a random process. In this sense of "random", you don't use
bits at all to determine if the number is random. You inspect the
source.

People doing crypto would never call a finite string "random", for
example, but only a finite piece of a random string. :-)

-- 
Darren New / Senior MTS & Free Radical / Invisible Worlds Inc.
       San Diego, CA, USA (PST).  Cryptokeys on demand.
     This is top-quality raw fish, the Rolls-Rice of Sushi!



More information about the Python-list mailing list