commons-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Gilles (JIRA)" <j...@apache.org>
Subject [jira] [Created] (RNG-20) Initial state diversity
Date Sat, 08 Oct 2016 13:58:20 GMT
Gilles created RNG-20:
-------------------------

             Summary: Initial state diversity
                 Key: RNG-20
                 URL: https://issues.apache.org/jira/browse/RNG-20
             Project: Commons RNG
          Issue Type: Improvement
            Reporter: Gilles
            Assignee: Gilles
            Priority: Minor
             Fix For: 1.0


A generator usually _assumes_ that its initial state is "sufficiently diverse" for it to perform
correctly.
For example, it is known that some algorithms (e.g. "Mersenne Twister") will be "stuck" by
the all-zero state.

To ensure diversity when the seed (passed by the user) contains less information than the
state can hold, {{SeedFactory}} provides a {{fillState}} method to be called from within the
{{setSeedInternal}} method of an RNG implementation.

In this ticket, it is proposed that the actual code (for replacing zeroes in the initial state)
is inspired from the one that already existed in class {{AbstractWell}}.

Note: neither the code in {{AbstractWell}} nor another code, in {{MersenneTwister}} and  {{MersenneTwister64}}
(with a similar purpose), will be replaced by a call to {{fillState}} because for those algorithms,
the filling procedure is part of the reference code (and is thus necessary in order to reproduce
the exact same output).




--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message