Posted by
Ferdinando M. Ametrano-2 on
URL: http://quantlib.414.s1.nabble.com/Ziggurat-algorithm-for-normal-deviates-tp13305p13306.html
On Tue, Jun 15, 2010 at 5:07 PM, Kakhkhor Abdijalilov
<
[hidden email]> wrote:
> In my tests I paired Mersenne
> Twister with Acklam's approximation, boost's inverse CDF and Ziggurat
> to generate standard normal distribution.
for those who wonder Acklam's approximation is the (best) inverse CDF
implemented in QuantLib
> Ziggurat method was 3 times
> faster than Acklam's approximation and 7 times faster than using
> boost's inverse CDF.
ok, so Ziggurat is interesting, even if it is only for pseudo-random
and not for quasi-random
> GSL version uses smaller look up
> table and exponential distribution for the tails.
smaller than what ? what is this lookup table ?
> I might create my
> own lookup table, but it would be exactly the same.
of course, if it is the inverse cdf lookup table it's no problem to
generate it again (even with Acklam improved by using one iteration of
Halley's rational method in order to obtain full machine precision)
> GSL is licensed under GPL v3, but I adopted only
> "matematical formula", the implementation follows the original
> Ziggurat paper. My implementation is hardwired to use Mersenne Twister
> directly and doesn't rely on GSL.
> [...] The math formula
> is from the published paper, not from GSL. Given that the only thing I
> took from GSL is precomputed look up table, is my implementation
> covered by GPL?
I would recompute the lookup table and then it's not covered by GPL
ciao -- Nando
------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the
lucky parental unit. See the prize list and enter to win:
http://p.sf.net/sfu/thinkgeek-promo_______________________________________________
QuantLib-dev mailing list
[hidden email]
https://lists.sourceforge.net/lists/listinfo/quantlib-dev