Problem with random seed
Locked
Problem with random seed
 
|
Hi again
There is a small problem with the random number generator. When a Rng is constructed, it initialises the generator with a seed based on the current time. However, (at least in VC++) the resolution of the timer is only one second. This means that the same sequence of random numbers will be generated. Normally, this is not a problem, as e.g. the PathGenerator next() method will not be reseeded with the same seed. But the problems shows, when PathGenerator is called repeatedly from an external program. My particular program is in the QuantLibMma extension, where the following code will give 5 identical paths. model = BlackScholesModel[100, 0.10, 0.01, 0.15]; Table[MonteCarloPath[model, 1, 20], {5}] Does anyone know how to solve this problem? Is there a seed with a higher resolution the QL_TIME? Cheers... Niels |
|
|
Hi,
Some quick ideas here: - use high-resolution performance counters (QueryPerformanceCounter) - use GUIDs (eg. CoCreateGuid) to get some 128 unique bits - use a master RNG to deliver seeds (use a different alg. here) - get a seed only once, and reseed with (++seed) The last one may be the most portable and efficient... Jens. -----Original Message----- From: [hidden email] [mailto:[hidden email]] On Behalf Of Niels Elken Sønderby Sent: Wednesday, May 14, 2003 10:49 PM To: QuantLib-dev Subject: [Quantlib-dev] Problem with random seed Hi again There is a small problem with the random number generator. When a Rng is constructed, it initialises the generator with a seed based on the current time. However, (at least in VC++) the resolution of the timer is only one second. This means that the same sequence of random numbers will be generated. Normally, this is not a problem, as e.g. the PathGenerator next() method will not be reseeded with the same seed. But the problems shows, when PathGenerator is called repeatedly from an external program. My particular program is in the QuantLibMma extension, where the following code will give 5 identical paths. model = BlackScholesModel[100, 0.10, 0.01, 0.15]; Table[MonteCarloPath[model, 1, 20], {5}] Does anyone know how to solve this problem? Is there a seed with a higher resolution the QL_TIME? Cheers... Niels ------------------------------------------------------- Enterprise Linux Forum Conference & Expo, June 4-6, 2003, Santa Clara The only event dedicated to issues related to Linux enterprise solutions www.enterpriselinuxforum.com _______________________________________________ Quantlib-dev mailing list [hidden email] https://lists.sourceforge.net/lists/listinfo/quantlib-dev |
RE: Problem with random seed
|
|
At 12:21 AM 5/15/03 +0200, Jens Thiel wrote:
>Some quick ideas here: > >- use high-resolution performance counters (QueryPerformanceCounter) >- use GUIDs (eg. CoCreateGuid) to get some 128 unique bits >- use a master RNG to deliver seeds (use a different alg. here) >- get a seed only once, and reseed with (++seed) Another quick one: - Use a global RNG, instead of creating one each time you call MonteCarloPath. Write another function to reinitialize it with a given seed if one wants to do so. This would also make your simulations more akin to what's usually done---i.e., create a RNG, and make paths by using successive draws from that one RNG. I don't think there's many people around reinitializing the generator for each path. Bye, Luigi |
|
|
>>>>> "Gigi" == Luigi Ballabio <[hidden email]> writes:
>> - use high-resolution performance counters >> (QueryPerformanceCounter) - use GUIDs (eg. CoCreateGuid) to get >> some 128 unique bits - use a master RNG to deliver seeds (use a >> different alg. here) - get a seed only once, and reseed with >> (++seed) Gigi> Another quick one: Gigi> - Use a global RNG, instead of creating one each time you Gigi> call MonteCarloPath. Write another function to reinitialize Gigi> it with a given seed if one wants to do so. Gigi> This would also make your simulations more akin to what's Gigi> usually done---i.e., create a RNG, and make paths by using Gigi> successive draws from that one RNG. I don't think there's Gigi> many people around reinitializing the generator for each Gigi> path. The problem of using different independent RNG is typical in parallel montecarlo applications, and using two different RNG algorithms (one for generating seeds and one for generating paths) is a known solution. One known approach is the "Lagged Fibonacci" method (try a search with google and you'll find some implementations), anyway you don't get generators with a huge period (I think something like 10E4 - 10E5) for the paths. On the other side, using only one generator for generating all the paths could not scale in the number of paths, anyway you could use a very high period RNG (the mersenne-twister, which have a huge period would probably fit financial applications needs). The Mersenne-Twister inventors released a mersenne-twister variant suitable for parallel montecarlo applications, anyway its code is GPL so, unfortunately, impossible to include in QL. I can't remember the authors' site URL (again, a google search should lead you there). I hope this helps, Enrico -- Enrico Sirola <[hidden email]> "Random generators should noy be chosen at random" - D. Knuth |
Re: Problem with random seed
|
|
Hi Enrico
>The Mersenne-Twister inventors released a mersenne-twister variant >suitable for parallel montecarlo applications, anyway its code is GPL >so, unfortunately, impossible to include in QL. I can't remember the >authors' site URL (again, a google search should lead you there). are you sure? I've just checked http://www.math.keio.ac.jp/matumoto/emt.html and since april 2001 the code of MT is licensed under BSD. Besides I miss the variant for parallel Monte Carlo. Any direction is appreciated ------------ ciao -- Nando PS this thread started here on quantlib-dev and it might be too late to switch to quantlib-users. anyway whenever possible I would appreciate if such discussion would happen on quantlib-users since i do trust a larger audience could contribute useful suggestions. Let's keep quantlib-dev as back-channel for discussion we don't want to be public, as confessing ignorance or difficulties with stochastic calculus ;-) |
|
|
for ql-users subscribers: this is a follow-up to a discussion going on
on ql-dev... It could be of broader interest so nando suggested to switch. here we go: >>>>> "Nando" == Ferdinando Ametrano <[hidden email]> writes: Nando> Hi Enrico >> The Mersenne-Twister inventors released a mersenne-twister >> variant suitable for parallel montecarlo applications, anyway >> its code is GPL so, unfortunately, impossible to include in >> QL. I can't remember the authors' site URL (again, a google >> search should lead you there). Nando> are you sure? I've just checked Nando> http://www.math.keio.ac.jp/matumoto/emt.html and since Nando> april 2001 the code of MT is licensed under BSD. Besides I Nando> miss the variant for parallel Monte Carlo. Any direction is Nando> appreciated Yes, MT is BSD, but their dynamic creation code is LGPL. Please see the content of the package at http://www.math.h.kyoto-u.ac.jp/~matumoto/RAND/DC/dc.html Nando> PS this thread started here on quantlib-dev and it might be Nando> too late to switch to quantlib-users. anyway whenever Nando> possible I would appreciate if such discussion would happen Nando> on quantlib-users since i do trust a larger audience could Nando> contribute useful suggestions. Let's keep quantlib-dev as Nando> back-channel for discussion we don't want to be public, as Nando> confessing ignorance or difficulties with stochastic Nando> calculus ;-) ok, this isn't related to stochastic calculus so i think i can crosspost ;-) ciao ciao enrico -- Enrico Sirola <[hidden email]> gpg public key available from www.keyserver.net, Key ID 0x377FE07F Key fingerprint = B446 7332 ED55 BC68 5FE8 DE0F 98DF EC86 377F E07F |
«
Return to quantlib-dev
|
1 view|%1 views
| Free forum by Nabble | Edit this page |