[R] Problem in SVM model generation

[David Winsemius]

On Oct 15, 2015, at 4:51 AM, Bhawana Sahu wrote:

> I am using R in my project for analysis of data, and I have generated SVM
> model using kernlab package with the dataset (3000 rows and 281 columns).
> Now I want to generate this model for dataset containing 8000 rows and 281
> column, but here I am getting an error saying that "cannot allocate vector
> of size 2.4 gb"
> 
> Earlier I was running this model on system with 4GB RAM, now I tried this
> on the system having RAM size of 64 gb, I tried all possible way to
> increase memory limit of R, and virtual memory, but problem persist.

The size of _contiguous_ memory is what's important and that will depend on the efficiency of your memory management (with Windows being accused of inferiority in the past) as well as how many other programs you have loaded and the degree of memory fragmentation. I'm not sure that you can resize memory when you already deep in a session. I thought it needed to be done early in the startup process but I'm not currently using Windows so am only reporting what I read in Rhelp

?'Memory-limits'
?Startup 

Delete any (possibly invisible) .RData file. Restart with a clean session of both your OS and R. You can get the size of objects currently in the R workspace with this function that I think I copied from one of Dirk Eddelbeuttel's or Bill Dunlap's posts:


getsizes <- 
function (num=10)  # change the num to different values to see more objects
{
    z <- sapply(ls(envir = globalenv()), function(x) object.size(get(x)))
    (tmp <- as.matrix(rev(sort(z))[1:num]))
}


getsizes()

> 
> Please suggest me what can be the issue with this, why am I getting this
> error, Is there any limitation of this package.
> 

Packages are not usually the culprit. It's usually program bloat by the user. I am usually guilty of having too many images and webpages open at the same time.


> Thank you
> 
> 	[[alternative HTML version deleted]]
> 
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> 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.

David Winsemius
Alameda, CA, USA