\name{CvLog}

\title{Cross validation Leave Out Group}
\description{
  Define a MlData partition in train/test sets.
  
  There is 2 way to construct a CvLog.
  If \code{trainProp} is provided then it build a balanced train/test sets.
  Else force the partition rows with testRows.
}
\usage{
##Constructor
CvLog(mlData=NULL, trainProp=80, trainRows=NULL, foldName=NULL
      cvType='LOG',foldName='GLOBAL'
       mlDataTrain=NULL,mlDataTest=NULL)

##Mis methods

setData(mlData)/resetData()
#Memory saving tricks should be replaced by clever MlData

## Misc accessors
vector   getTrainRows()
vector   getTestRows()

run(clf)
#clf : a set of classifiers.
#run classifier(s):
#learn on train set, predict on train and test set.
#The output is a CvResults.

}
\arguments{
  \item{mlData}{\code{MlData} object} 
  \item{trainProp}{the percentage of training subjects, is used to
    randomly build balanced train/test set if \code{trainRows} is omitted.} 
  \item{testRows}{Manually set the the test rows}
  \item{mlDataTrain}{Manually set a \code{MlData} object for train data}
  \item{mlDataTest} {Manually set a \code{MlData} object for test data}
  \item{foldName} {Optionnal if one want to give a name this output
    (Used by \code{CvLoo} and \code{CvNFolds} in order to identify
    distinct runs)} 
  \item{clf}{\code{Clf} "classifier" Object}
}

\value{
  \item{CvResults}{Cross validation output}
}

\seealso{
  \link{\code{MlData}} \link{\code{CvResults}}
}
\examples{
mlData=MlData(file='small.csv',grpCol='class',subjCol='subject')
cv=CvLog(mlData)
resLda=cv$run(clf=ClfLda())
resSvm=cv$run(clf=ClfSvm())

resLda$testAccuracy()
resSvm$testAccuracy()

##Everything on a single line:
res=CvLog(mlData)$run(clf=ClfLda())
res$getData()
}