[
https://issues.apache.org/jira/browse/MATH-1314?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15091112#comment-15091112
]
Luc Maisonobe commented on MATH-1314:
-------------------------------------
Nothing can ensure *all* seed are good enough. For any procedure, there will be
corner cases that may be poor.
The procedures that we have now were provided by the algorithm creators, we
cannot do anything more
than trusting these experts. I don't think we can provide any tools. The
generators are good, and they
do create an entropy pool efficiently. Checking the very slight non-uniformity
of these pool is beyond
our capabilities. We could do that if the generators were poor, but they are
good.
Concerning the performances, they we obtained on my machine, and its
characteristics are shown,
because as usual the results depend on the computer. They are not expected to
be really accurate,
only to give some rough order of magnitudes. Also as the implementations were
changed since the
benchmarks have been run, I am not surprised these rough order of magnitudes
are not true
anymore. So you can either remove the table or update it. Is the
MersenneTwister much faster or
only marginally faster?
> RNG: Warn users about "seeding"
> -------------------------------
>
> Key: MATH-1314
> URL: https://issues.apache.org/jira/browse/MATH-1314
> Project: Commons Math
> Issue Type: Wish
> Reporter: Gilles
> Labels: doc
> Fix For: 4.0
>
>
> The "package-info.java" file of {{o.a.c.m.random}} does not mention the
> problem of seeding.
> Many users of CM could not be aware that it is not sufficient to "randomly"
> choose a seed in order to ensure a random sequence.
> I think that this is what is illustrated by random failures of some unit
> tests (when the seed is "randomly" selected).
> Do the intricate initialization procedures provided in some implementations
> (WELL family and ISAAC) ensure that all seeds are good enough?
> Should we provide some tool to test a seed?
> By the way, the WELL performances listed on [this
> table|http://commons.apache.org/proper/commons-math/javadocs/api-3.6/org/apache/commons/math3/random/package-summary.html]
> do not correspond to the results obtained on my machine with our
> {{PerfTestUtils}} benchmark: the {{MersenneTwister}} is invariably faster
> than all WELL implementations.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
