Boost randon number generator
toeklk opened this issue · comments
migrated from Bugzilla #491
status NEW severity normal in component wrappers for 0.7.0
Reported in version trunk on platform All
Assigned to: BFL mailinglist
Original attachment names and IDs:
- boost_rng_seed.patch (ID 233)
On 2008-01-18 13:27:55 +0100, Tinne De Laet wrote:
Apparently, calling sampleFrom on a Gaussian always results in the same set of samples when using Boost as a random number generator. It appears like boost has a fixed list of random samples, resulting in the same samples when executing the program multiple times. This problem does not occur when using lti as a randon number generator. Is this an intrinsic problem with Boost (or are we not using the rng properly)? I found the following note on http://www.boost.org/libs/random/random-generators.html "Note: These random number generators are not intended for use in applications where non-deterministic random numbers are required. See nondet_random.html for a choice of (hopefully) non-deterministic random number generators." So maybe we should use the proposed class when we want non-deterministic randon numbers. Tinne
On 2008-01-18 17:06:35 +0100, Tinne De Laet wrote:
This bug causes the new pdf_test (see BUG # 441, Attachment # 214) to fail deterministically. Apparently the 496'th sample is located outside the 4 sigma region. Tinne
On 2008-02-09 20:13:47 +0100, Klaas Gadeyne wrote:
Created attachment 233 quick-n-dirty fix for the test This patch show that this can be fixed using a seed. However, this is only a quick and dirty fix, so I'm not in favour of having it applied. The proper way to fix this is to make objects of the RNGs