New random generator
Locked
 
|
Hi guys
Konstantin who is a biologist developed for a "bacteria" eating monster simulation system, two new random number generators. I think that this is great because we need more random generators :) - Combined Random Number Generator is based on the algorithm described by PIERRE L'ECUYER in "Efficient and Portable Combined Random Number Generators" [Communications of the ACM, June 19, Volume 31, Number 6, pp. 742-749, references p.774]. - ExpInvConGen is an explicit inversive congruential generator, constructed according to "Good random number generators are (not so) easy to find" by P. Hellekalek (1998) and extended euclidean algorithm. I plan to package them and publish them in SciTalk with their own configuration. But I wanted to know what you think. Stef |
|
|
On Thu, Mar 19, 2015 at 9:36 PM, stepharo <[hidden email]> wrote:
> Hi guys > > Konstantin who is a biologist developed for a "bacteria" eating monster > simulation system, two new random number generators. > I think that this is great because we need more random generators :) > > - Combined Random Number Generator is based on the algorithm described by > PIERRE L'ECUYER in "Efficient and Portable Combined Random Number > Generators" [Communications of the ACM, June 19, Volume 31, Number 6, pp. > 742-749, references p.774]. > > - ExpInvConGen is an explicit inversive congruential generator, constructed > according to "Good random number generators are (not so) easy to find" by P. > Hellekalek (1998) and extended euclidean algorithm. > > I plan to package them and publish them in SciTalk with their own > configuration. But I wanted to know what you think. Great ! Maybe better to integrate them in Sci-talk with the existing Random Number Generator (RNG). We have already 5 RNGs. I'm working on refactoring the RNG and distributions in Sci-Talk, because there is some redundancy until now: https://groups.google.com/forum/#!topic/scismalltalk/zXEyeGvTmJk Regards, -- Serge Stinckwich UCBN & UMI UMMISCO 209 (IRD/UPMC) Every DSL ends up being Smalltalk http://www.doesnotunderstand.org/ |
|
|
the method from l'Ecuyer is already used in the texas instruments graphic calculators. But they don't seem to have patented it (I don't know how to check that). Anyway I agree to have more RNGs which can be useful for statistics simulations or kryptography. And I would like to know how one can play with the bacteria ;-) Alain PS did anyone receive the mail that I sent yesterday about syllogisms? On Fri, Mar 20, 2015 at 12:43 AM, Serge Stinckwich <[hidden email]> wrote:
|
|
|
:) I will tell him.
|
|
|
In reply to this post by SergeStinckwich
Serge
I would like to have a package with Random-Number generators with its own configuration and without the distributions. Do you plan to package them separately? Because we often need randomnumber generator alone. Stef Le 19/3/15 21:43, Serge Stinckwich a écrit : > On Thu, Mar 19, 2015 at 9:36 PM, stepharo <[hidden email]> wrote: >> Hi guys >> >> Konstantin who is a biologist developed for a "bacteria" eating monster >> simulation system, two new random number generators. >> I think that this is great because we need more random generators :) >> >> - Combined Random Number Generator is based on the algorithm described by >> PIERRE L'ECUYER in "Efficient and Portable Combined Random Number >> Generators" [Communications of the ACM, June 19, Volume 31, Number 6, pp. >> 742-749, references p.774]. >> >> - ExpInvConGen is an explicit inversive congruential generator, constructed >> according to "Good random number generators are (not so) easy to find" by P. >> Hellekalek (1998) and extended euclidean algorithm. >> >> I plan to package them and publish them in SciTalk with their own >> configuration. But I wanted to know what you think. > Great ! > > Maybe better to integrate them in Sci-talk with the existing Random > Number Generator (RNG). We have already 5 RNGs. > I'm working on refactoring the RNG and distributions in Sci-Talk, > because there is some redundancy until now: > https://groups.google.com/forum/#!topic/scismalltalk/zXEyeGvTmJk > > Regards, > |
|
|
In reply to this post by stepharo
Ok I added the two RNG in Math-Random.
I cleaned also the hierarchy. - no need to have initialize as abstract to force all the subclass to define an empty method. Strange design - tab a bit the code and comments. I produced a new configuration based on SergeStinkwich 20 (which was not in the latest configurationOf). Now I saw that the postload invoke this method but DhbMaximumLikelihoodHistogramFit does not exist. So the postload should be removed. aliasToTypoClassNameDhbMaximumLikekihoodHistogramFit Smalltalk at: #DhbMaximumLikekihoodHistogramFit put: DhbMaximumLikelihoodHistogramFit |
|
|
On Fri, Mar 20, 2015 at 9:32 AM, stepharo <[hidden email]> wrote:
> Ok I added the two RNG in Math-Random. > I cleaned also the hierarchy. > - no need to have initialize as abstract to force all the subclass to > define an empty method. Strange design > - tab a bit the code and comments. > > I produced a new configuration based on SergeStinkwich 20 (which was not in > the latest configurationOf). > > > Now I saw that the postload invoke this method but > DhbMaximumLikelihoodHistogramFit does not exist. > So the postload should be removed. > > > aliasToTypoClassNameDhbMaximumLikekihoodHistogramFit > > Smalltalk at: #DhbMaximumLikekihoodHistogramFit put: > DhbMaximumLikelihoodHistogramFit > Thank you Stephane ! I will have a look to your modifications. At the moment, distributions and RNG are little messy. There is some distributions/RNGs in Math-Random, others in DHB packages. The API is not really the same between them. I have to clean everything in order to have a clean and standard API. -- Serge Stinckwich UCBN & UMI UMMISCO 209 (IRD/UPMC) Every DSL ends up being Smalltalk http://www.doesnotunderstand.org/ |
|
|
In reply to this post by stepharo
On 03/20/2015 09:32 AM, stepharo wrote:
> Now I saw that the postload invoke this method but > DhbMaximumLikelihoodHistogramFit does not exist. > So the postload should be removed. Hi Stephane, hi Serge, this postload was made by someone as a kind of improvised depreciation (DhbMaximumLikelihoodHistogramFit does exist). but it certainly is not necessary anymore. i can update the config accordingly, if thats ok with you. i would keep the #alias... method nevertheless so that the older versions still work. and i noticed that you did not change the stable version for pharo3 to the new version, but i could not find any incompatibilities in the new rngs with 3.0. i guess i could change the stable version for 3.0 too, is that correct? werner |
|
|
Le 20/3/15 12:29, Werner Kassens a écrit : > On 03/20/2015 09:32 AM, stepharo wrote: >> Now I saw that the postload invoke this method but >> DhbMaximumLikelihoodHistogramFit does not exist. >> So the postload should be removed. > > Hi Stephane, hi Serge, > this postload was made by someone as a kind of improvised depreciation > (DhbMaximumLikelihoodHistogramFit does exist). but it certainly is not > necessary anymore. i can update the config accordingly, if thats ok > with you. i would keep the #alias... method nevertheless so that the > older versions still work. and i noticed that you did not change the > stable version for pharo3 to the new version, but i could not find any > incompatibilities in the new rngs with 3.0. i guess i could change the > stable version for 3.0 too, is that correct? I did not change it because I did not test it on 30. > werner > > > |
«
Return to Pharo Smalltalk Users
|
1 view|%1 views
| Free forum by Nabble | Edit this page |