random number generators

tibbar · February 15th, 2008, 11:49 pm

Which is the best algorithm? GSL seems to suggest ranlux, others say it's ran2 from numerical recipes... Some convergence testing I did seems to suggest ranlux is a lot better.

Stylz · February 18th, 2008, 7:42 pm

mersenne twister is the best quasi-random out there IMHO

ExSan · February 18th, 2008, 7:59 pm

"Which is the best algorithm? GSL seems to suggest ranlux, others say it's ran2 from numerical recipes... Some convergence testing I did seems to suggest ranlux is a lot better. "ExSan uses core code from Numerical Recipes in C++ 2nd Ed. You may use the attached program that generates random numbers for your final evaluation.MENU: 1 Generate Random NumbersRAND MENU1 Uniform Random Deviate between 0.0 and 1,2 Normally distributed deviate N(0, 1) ,3 Exponential,4 Pareto,5 Gamma,6 Poisson Deviates, 7 Correlated Random Numbers,8 Log Normal Distributed Random number,9 Fisher Distributed Random Numbers,10 Hyperbolic Distributed Random Numbers ...men at work,b Simulating Geometric Brownian Motion, !NEWo Simulating Ornstein-Uhlenbeck Process, !NEWv Simulating Interest Rates with the Vasicek Model, !NEWs Simulating the Schwartz Type 1 Stochastic Process, !NEWFrom the download-able file extract executable EXSAN to your desktop.The very first time EXSAN is executed it will create the default input/output directory c:\exsanIn order to access this directory START / RUN and type in c:\exsan

Cuchulainn · February 18th, 2008, 8:08 pm

QuoteOriginally posted by: Stylzmersenne twister is the best quasi-random out there IMHOThat's the one, for the moment. (sequential C++)

AVt · February 18th, 2008, 8:19 pm

and to be used for what?

tibbar · February 18th, 2008, 9:42 pm

Mainly risk management Monte Carlo models (large models, lots of heavy tailed distributions with correlation). Interested in percentiles and conditional tail expectation of various outputs.robsan7 - that's the ran2 I've already tested - seems to slower poor convergence compared to ranlux.I'm trying the twister one at moment. Looks good so far. Thanks guys.

Zedr0n · February 19th, 2008, 5:21 am

I've kinda thought that for Monte Carlo low discrepancy numbers are better than quasi-random, no?

tibbar · February 19th, 2008, 8:04 am

You mean quasi are better than pseudo ;-)Of course, but when the dimensionality gets high, you need a back up plan...

Zedr0n · February 19th, 2008, 2:38 pm

Yes, sure hm, mersenne twister is definitely pseudo-random then, not quasi-random.

ExSan · April 20th, 2012, 10:28 am

This website offers true random numbers to anyone on the internet. The random numbers are generated in real-time in lab by measuring the quantum fluctuations of the vacuumQuantum Random Numbers

Traden4Alpha · April 20th, 2012, 1:20 pm

QuoteOriginally posted by: ExSanThis website offers true random numbers to anyone on the internet. The random numbers are generated in real-time in lab by measuring the quantum fluctuations of the vacuumQuantum Random NumbersI sincerely hope that they don't let two people use the same numbers or the long-distance correlations could get quite spooky.

Cuchulainn · April 20th, 2012, 1:30 pm

and here

animeshsaxena · April 21st, 2012, 9:42 am

and no one said sobol