reproducible-caret-models.R

# reproducible caret models
# http://stackoverflow.com/questions/13403427/fully-reproducible-parallel-models-using-caret


library(doParallel); library(caret)

#create a list of seed, here change the seed for each resampling
set.seed(123)
seeds <- vector(mode = "list", length = 11)#length is = (n_repeats*nresampling)+1
for(i in 1:10) seeds[[i]]<- sample.int(n=1000, 3) #(3 is the number of tuning parameter, mtry for rf, here equal to ncol(iris)-2)

seeds[[11]]<-sample.int(1000, 1)#for the last model

 #control list
 myControl <- trainControl(method='cv', seeds=seeds, index=createFolds(iris$Species))

 #run model in parallel
 cl <- makeCluster(detectCores())
 registerDoParallel(cl)
 model1 <- train(Species~., iris, method='rf', trControl=myControl)

 model2 <- train(Species~., iris, method='rf', trControl=myControl)
 stopCluster(cl)

 #compare
 all.equal(predict(model1, type='prob'), predict(model2, type='prob'))
[1] TRUE