[BioC] problems using the latest version of e1071 (svm)

[Javier Pérez Florido]

Dear list,
For R version 2.13.0, I had no problems running the following code (CMA 
package was used to load golub data, but detach afterwards trying to 
isolate the problem):

library(CMA)
data(golub)
detach(package:CMA)
library(e1071)

golubY <- golub[,1]
golubX <- as.matrix(golub[,-1])
ratio <- 2/3
golubY2 <- as.factor(golubY)
levels(golubY2) <- 1:nlevels(golubY2)
golubY2<-as.numeric(golubY2)-1

set.seed(777)
learnind <- sample(length(golubY), size=floor(ratio*length(golubY)))

Xtrain=golubX[learnind,]
Ytrain=as.factor(golubY2[learnind])

svm.model <- 
svm(x=Xtrain,y=Ytrain,seed=set.seed(321),cost=1,kernel="linear",type="C-classification")

However, for R version 2.14.2 and running the same code, R was closed by 
Windows because a problem in svm function. Here is the session info:

R version 2.14.2 (2012-02-29)
Platform: x86_64-pc-mingw32/x64 (64-bit)

locale:
[1] LC_COLLATE=Spanish_Spain.1252  LC_CTYPE=Spanish_Spain.1252
[3] LC_MONETARY=Spanish_Spain.1252 LC_NUMERIC=C
[5] LC_TIME=Spanish_Spain.1252

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] e1071_1.6      class_7.3-3    Biobase_2.14.0

loaded via a namespace (and not attached):
[1] CMA_1.12.0   tools_2.14.2

%%%%%%%%%%%%%%

I've tried in Linux O.S. too, obtaining the following error:


  *** caught segfault ***
address (nil), cause 'memory not mapped'

Traceback:
  1: .C("svmtrain", as.double(if (sparse) x at ra else t(x)), 
as.integer(nr),     as
.integer(ncol(x)), as.double(y), as.integer(if (sparse) x at ia else 
0),     as.int
eger(if (sparse) x at ja else 0), as.integer(type), as.integer(kernel),     
as.inte
ger(degree), as.double(gamma), as.double(coef0), as.double(cost),     
as.double(
nu), as.integer(weightlabels), as.double(class.weights),     
as.integer(length(c
lass.weights)), as.double(cachesize),     as.double(tolerance), 
as.double(epsilo
n), as.integer(shrinking),     as.integer(cross), as.integer(sparse), 
as.integer
(probability),     as.integer(seed), nclasses = integer(1), nr = integer(1),
  index = integer(nr), labels = integer(nclass), nSV = 
integer(nclass),     rho =
  double(nclass * (nclass - 1)/2), coefs = double(nr *         (nclass - 
1)), sig
ma = double(1), probA = double(nclass *         (nclass - 1)/2), probB = 
double(
nclass * (nclass - 1)/2),     cresults = double(cross), ctotal1 = 
double(1), cto
tal2 = double(1),     error = err, PACKAGE = "e1071")
  2: svm.default(x = Xtrain, y = Ytrain, seed = set.seed(321), cost = 
1,     kern
el = "linear", type = "C-classification")
  3: svm(x = Xtrain, y = Ytrain, seed = set.seed(321), cost = 1, kernel 
= "linear
",     type = "C-classification")

Possible actions:
1: abort (with core dump, if enabled)
2: normal R exit
3: exit R without saving workspace
4: exit R saving workspace
Selection:

R version 2.14.2 (2012-02-29)
Platform: x86_64-unknown-linux-gnu (64-bit)

locale:
  [1] LC_CTYPE=en_US.iso885915       LC_NUMERIC=C
  [3] LC_TIME=en_US.iso885915        LC_COLLATE=en_US.iso885915
  [5] LC_MONETARY=en_US.iso885915    LC_MESSAGES=en_US.iso885915
  [7] LC_PAPER=C                     LC_NAME=C
  [9] LC_ADDRESS=C                   LC_TELEPHONE=C
[11] LC_MEASUREMENT=en_US.iso885915 LC_IDENTIFICATION=C

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base

other attached packages:
[1] e1071_1.6      class_7.3-3    Biobase_2.14.0

loaded via a namespace (and not attached):
[1] CMA_1.12.0   tools_2.14.2

Thanks,
Javier