About the efficiency of R optimization function
You might get some information about what is available in R using
the new "RSiteSearch" package. For example:
library(RSiteSearch)
ipm <- RSiteSearch.function('interior point method')
hits(ipm) # 39
SQP <- RSiteSearch.function('SQP')
hits(SQP) # 2
sqp <- RSiteSearch.function('sequential quadratic programming')
hits(sqp) # 2
sqp. <- (sqp | SQP)
dim(sqp.) # 4 10
ipm.. <- (ipm | sqp.)
HTML(ipm..)
NOTE: The "|" = "unionRSiteSearch" function is available in
version 1.0-3, which was released earlier today and should arrive at
your favorite CRAN mirror tomorrow or Sunday. Or you can get it now via
'install.packages("RSiteSearch", repos="http://R-Forge.R-project.org")'.
Hope this helps.
Spencer Graves
popo UBC wrote:
Hi Charlie, Thank you so much for suggestions!! Actually, I used the optimization toolbox in MABLAB before and I even wrote some numerical optimization programs by myself. As far as I know, some commercial optimization softwares had already replaced L-BFGS-B by more advanced algorithms, such as interior point method, SQP(sequential quadratic programming), implemented under trust region strategy. So, - Have you ever tried these techniques? Are they available in R already? - In your previous experieces, did R work satisfactory? I mean, was it often that R failed to converge or spent too much time? - Mainly, I need to calculate the MLE. But I really have no idea what the likelihood may looks like. According to your experiences, would the likelihood function be too complicated? Is L-BFGS-B good enough? Thanks again!! Popo 2009/5/14 cls59 <sharpsteen at mac.com>
popo UBC wrote:
Hi all!
The objective function I want to minimize contains about 10 to 20
variables,
maybe more in the future. I never solved such problems in R, so I had no
idea about the efficiency of R's optimization functions. I know doing
loop
in R is quite slow, so I am not sure whether this shortage influences the
speed of R's optimization functions.
I would be very appreciated if anyone could share some experiences with
me.
The speed, stability of the R's optimization functions. Is it helpful to
call a C/Fortran code to do the job, if possible.
Many thanks in advance.
Popo
Many functions available in R are implemented using a compiled language such as C or Fortran- not the R language it's self. For example, the "Source" section of the help page for optim states that the code for the Nelder-Mead, BFGS and Conjugate Gradient methods were translated to C from Pascal and then further optimized. The L-BFGS-B method appears to be implemented as Fortran code. Looking at the source of the optim function reveals that results are computed by a call to .Internal(). Such calls usually indicate that R is handing computations off to a compiled, rather than interpreted, routine. If you have C or Fortran code you would prefer to use, take a look at the help pages for .C() and .Fortran() as well as the "Writing R Extensions" manual. The command line tool R CMD SHLIB will help you compile your code to shared libraries that can be loaded by R using dyn.load(). -Charlie ----- Charlie Sharpsteen Undergraduate Environmental Resources Engineering Humboldt State University -- View this message in context: http://www.nabble.com/About-the-efficiency-of-R-optimization-function-tp23552061p23552668.html Sent from the R help mailing list archive at Nabble.com.
______________________________________________ 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<http://www.r-project.org/posting-guide.html> and provide commented, minimal, self-contained, reproducible code.
[[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.