A user is trying to run a parallel kNN program on R but he gets this error: Error in { : task 1 failed - "could not find function "knn""

library(class)

library(doSNOW)

library(foreach)

train <- read.csv('train.csv')

test <- read.csv('test.csv')

trainY <- read.csv('trainY.csv')

cl <- as.vector(as.matrix(trainY))

system.time(summary(knn(train, test, cl, k=25, prob = TRUE)))

clus <- makeCluster(4)

registerDoSNOW(clus)

countrows=nrow(test)

system.time(foreach( icount(countrows) ) %dopar% {

  summary(knn(train, test, cl, k=25, prob = TRUE))

})

stopCluster(clus)

In such case, we need to call library(class) on each of the nodes. foreach makes this easy via the .packages argument.

system.time(foreach( icount(countrows), .packages="class" ) %dopar% {

  summary(knn(train, test, cl, k=25, prob = TRUE))

})

Also we might need to export train, test and cl.

system.time(

  foreach( icount(countrows), .packages="class",

           .export=c("train","test","cl") ) %dopar% {

    summary(knn(train, test, cl, k=25, prob = TRUE))

  }

)