Random number generators
in [Fortran]
|
Prev: pointer and allocatable arrays
Next: f95 to windows dll
From: nmm1 on 23 Nov 2009 08:33 In article <yQuOm.56946$ze1.52594(a)news-server.bigpond.net.au>, robin <robin_v(a)bigpond.com> wrote: >| >| >George Marsaglia's KISS RNG has a period greater than >| >10**35, which is far far far greater than the 10**6 figure >| >that you suggest. >| >| Sigh. I am not talking about the period, but the precision, and >| the discreteness starts to become a serious problem if any one >| simulation uses more than about a million numbers. > >Your claims are vague and are unsubstantiated. >You have not said anything about about Marsaglia's specific >32-bit and 64-bit generators. >Nor have you suppled any information about the 32-bit >generators that you claim to have noticed. Do a literature search. I published the analysis. Regards, Nick Maclaren.
From: Harold Stevens on 23 Nov 2009 10:15 In <hee304$7r9$1(a)soup.linux.pwf.cam.ac.uk> Nick: [Snip...] > Do a literature search. I published the analysis. I admire your cheerful optimism where "robin" is involved. :) -- Regards, Weird (Harold Stevens) * IMPORTANT EMAIL INFO FOLLOWS * Pardon any bogus email addresses (wookie) in place for spambots. Really, it's (wyrd) at airmail, dotted with net. DO NOT SPAM IT. I toss GoogleGroup (http://twovoyagers.com/improve-usenet.org/).
From: steve on 23 Nov 2009 11:08 On Nov 23, 5:33 am, n...(a)cam.ac.uk wrote: > In article <yQuOm.56946$ze1.52...(a)news-server.bigpond.net.au>, > > robin <robi...(a)bigpond.com> wrote: > >| > >| >George Marsaglia's KISS RNG has a period greater than > >| >10**35, which is far far far greater than the 10**6 figure > >| >that you suggest. > >| > >| Sigh. I am not talking about the period, but the precision, and > >| the discreteness starts to become a serious problem if any one > >| simulation uses more than about a million numbers. > > >Your claims are vague and are unsubstantiated. > >You have not said anything about about Marsaglia's specific > >32-bit and 64-bit generators. > >Nor have you suppled any information about the 32-bit > >generators that you claim to have noticed. > > Do a literature search. I published the analysis. > > Regards, > Nick Maclaren. Is it this paper? Title: Cryptographic pseudo-random numbers in simulation Author: Maclaren, N. Source: Cambridge Security Workshop. Fast Software Encryption. 1994. p. 185-90 It's unavailable from the Univ. of Washington library? Do you have a pdf that is accessible? -- steve
From: nmm1 on 23 Nov 2009 11:31 In article <168fdf4b-0398-451f-915a-996d550afbce(a)b36g2000prf.googlegroups.com>, steve <kargls(a)comcast.net> wrote: >> >| >> >| Sigh. =A0I am not talking about the period, but the precision, and >> >| the discreteness starts to become a serious problem if any one >> >| simulation uses more than about a million numbers. >> >> >Your claims are vague and are unsubstantiated. >> >You have not said anything about about Marsaglia's specific >> >32-bit and 64-bit generators. >> >Nor have you suppled any information about the 32-bit >> >generators that you claim to have noticed. >> >> Do a literature search. =A0I published the analysis. > >Is it this paper? > >Title: Cryptographic pseudo-random numbers in simulation >Author: Maclaren, N. >Source: Cambridge Security Workshop. Fast Software Encryption. > 1994. p. 185-90 No. It's this one: A Limit on the Usable Length of a Pseudorandom Number Sequence (Journal of Statistical Computation and Simulation (1992), vol.\ 42, pp.\ 47-54). Regrettably, I have only offprints, and I managed to bury them in a previous office move. SOMEWHERE I will have the source of the paper! I will take a look tonight. Regards, Nick Maclaren.
From: Ron Shepard on 23 Nov 2009 12:38
In article <heeddb$h6f$1(a)soup.linux.pwf.cam.ac.uk>, nmm1(a)cam.ac.uk wrote: > A Limit on the Usable Length of a Pseudorandom Number Sequence (Journal of > Statistical Computation and Simulation (1992), vol.\ 42, pp.\ 47-54). I have no doubt that the original statement in this discussion, that less than a million numbers should be sampled from a 32-bit PRNG, is true for many statistical simulations. However, there are indeed many applications that are just fine with relatively short cycles. For example, is the simulation depends not only on the current state but on the history, then this can effectively increases the cycle length. Here is an example of this. Suppose that you want to generate a few random vectors of a million real elements, and supposed your method is successful as long as the vectors are linearly independent. In this case, your code will work correctly with short PRNG cycles (even those shorter than a million in length in some cases) that repeat during this process *unless* the cycle begins to repeat for the same vector index for two of the vectors. The odds of that happening are one out of a million, so effectively, the history of the sequences, which determines the index at which the cycle begins to repeat, works to eliminate the problems with the short cycle. This is not an obscure application of PRNGs, this is one of the more common ones. $.02 -Ron Shepard |