[BioC] Re: Re: segmentation error in pamr.knnimpute

Alice Le Bars alice.le-bars at nantes.inserm.fr
Mon May 9 11:54:07 CEST 2005


Hi Seth, 

My computer is a 2,67Ghz with 1Go RAM and 2,35Go of memory in C:/ driver (if R use the driver for his temporary files management), but I have our linux server to avoid this case (a lake of memory), he has a quadriprocessor and I think enough resources. 
I am using the R version 2.0.1  and PamR version 1.24 (the same for my computer and the server)
This is my loop :

for (i in 1: nbclas) {
    liste2 <- listdeuz[[i]]
    matclass <- liste2$mat.expression
    print(matclass[1:10,]  )
    matclassknn[[i]] <- f.kNN(matclass)
    }

nbclas = 2 
The tables of expression are included in a list named listdeuz,  liste2$mat.expression = 4* 1337 and 4*1327.
In my computer the program breakes without printing the 10 first lines , on the linux server he prints the first 10 lines. Those tables are a partition of my biggest table which could be impute by pmar.knnimpute. 
I also used knn excel's SAM four times to be sure of my two little tables and it worked everytimes. 

If you wish, I can send you the two tables.

Thanks for everything,
Alice. 

/Hi Alice,/

/Alice Le Bars <alice.le-bars at nantes.inserm.fr> writes:
/

>/> I am in trouble with pamr.knnimpute (library pamr) When I use this
>> function in a "for" loop with matrix of 4 samples at each loop, R
>> does a segmentation fault which ends with a fatal error (R close
>> itself) I used this function previously with 8 samples and 1327
>> genes and it worked.
>/
>
/
In order for us to try and help, please tell us:

- Your version of R and version of pamr
- Show us the code that is breaking (the for loop), if possible
- Tell us about the resources available on the machines you've tried
  this on (how much memory is the important thing).

The segfaulting you are reporting sounds like out of memory behavior,
but could also be a bug in pamr.  But you'll need to help of with much
more detail...

+ seth/

Alice Le Bars
INSERM U533 - IFR 26
1 Rue Gaston Veil- BP 53508
44 035 NANTES
FRANCE

Tel : (33) 2 40 41 29 86
Fax : (33) 2 40 41 29 50

http://cardioserve.nantes.inserm.fr/ptf-puce



More information about the Bioconductor mailing list