R help - Aug 2012 - random number generator with SNOW/ Parallel/ foreach

Dear All,

I have three classes of questions about generating random numbers with
different packages (windows xp 32bit R).
.
1. Suppose I would like to use package *foreach*, can I use current
Sys.time as a seed?
Although I can get the time up to1e-6 second precesion, the code below dose
not work well on a local machine with two cores. #################
library(foreach)
library(snow)
library(doSNOW)
testfun=function(){
             options(digits.secs=6)
             set.seed(slave.seed <- substring(Sys.time(), 21, 26))
             a <- rnorm(5,0,1)
             return(a)
                               }
maxkk <- 5
cl <- makeCluster(5, type = "SOCK")
registerDoSNOW(cl)
results=foreach (i = 1:maxkk, .combine = 'rbind') %dopar% testfun()
results
stopCluster(cl)
#################

2. Suppose I want to use *snow*, according to some reference, I need to use
*rsprng* as a generator. Does this generator relay on seed as well? But it
seems to me that if I do not use rsprng, it is also fine.
## code without rsprng ##
testfun <- function(){
   a <- rnorm(5,0,1)
   return(a)
       }
library(snow)
cl <- makeCluster(5,type="SOCK")
clusterCall(cl, testfun)
stopCluster(cl)

## code with rsprng ##
testfun <- function(){
   a <- rnorm(5,0,1)
   return(a)
       }
library(snow)
library(rsprng)
cl <- makeCluster(5,type="SOCK")
clusterSetupRNG(cl)
clusterCall(cl, testfun)
stopCluster(cl)

3. Suppose I would like to use the package *parallel, *do I still need *
rsprng? *Should I specify any seed in this case?

testfun <- function(){
   a <- rnorm(5,0,1)
   return(a)
       }
library(parallel)
cl <- makeCluster(5,type="SOCK")
clusterCall(cl, testfun)
stopCluster(cl)

My questions can also be viewed as these aspects: how do these generators
work with/without seed, what is the relation between these packages, what
is the best option of seed if needed.
Thank you very much for your help.

Best wishes,
Jie

	[[alternative HTML version deleted]]

I believe this is discussed in the vignette (extended documentation)
of the parallel package, in particular section 6: type

vignette("parallel")

to see. I've also attached a copy.

Michael

On Tue, Aug 7, 2012 at 7:57 PM, Jie <jimmycloud at gmail.com>
wrote:
> Dear All,
>
> I have three classes of questions about generating random numbers with
> different packages (windows xp 32bit R).
> .
> 1. Suppose I would like to use package *foreach*, can I use current
> Sys.time as a seed?
> Although I can get the time up to1e-6 second precesion, the code below dose
> not work well on a local machine with two cores. #################
> library(foreach)
> library(snow)
> library(doSNOW)
> testfun=function(){
>              options(digits.secs=6)
>              set.seed(slave.seed <- substring(Sys.time(), 21, 26))
>              a <- rnorm(5,0,1)
>              return(a)
>                                }
> maxkk <- 5
> cl <- makeCluster(5, type = "SOCK")
> registerDoSNOW(cl)
> results=foreach (i = 1:maxkk, .combine = 'rbind') %dopar% testfun()
> results
> stopCluster(cl)
> #################
>
> 2. Suppose I want to use *snow*, according to some reference, I need to use
> *rsprng* as a generator. Does this generator relay on seed as well? But it
> seems to me that if I do not use rsprng, it is also fine.
> ## code without rsprng ##
> testfun <- function(){
>    a <- rnorm(5,0,1)
>    return(a)
>        }
> library(snow)
> cl <- makeCluster(5,type="SOCK")
> clusterCall(cl, testfun)
> stopCluster(cl)
>
> ## code with rsprng ##
> testfun <- function(){
>    a <- rnorm(5,0,1)
>    return(a)
>        }
> library(snow)
> library(rsprng)
> cl <- makeCluster(5,type="SOCK")
> clusterSetupRNG(cl)
> clusterCall(cl, testfun)
> stopCluster(cl)
>
> 3. Suppose I would like to use the package *parallel, *do I still need *
> rsprng? *Should I specify any seed in this case?
>
> testfun <- function(){
>    a <- rnorm(5,0,1)
>    return(a)
>        }
> library(parallel)
> cl <- makeCluster(5,type="SOCK")
> clusterCall(cl, testfun)
> stopCluster(cl)
>
> My questions can also be viewed as these aspects: how do these generators
> work with/without seed, what is the relation between these packages, what
> is the best option of seed if needed.
> Thank you very much for your help.
>
> Best wishes,
> Jie
>
>         [[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide
http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: parallel.pdf
Type: application/pdf
Size: 292168 bytes
Desc: not available
URL:
<https://stat.ethz.ch/pipermail/r-help/attachments/20120807/5ffe785b/attachment.pdf>