CODEWITHSUNDEEP
rforeachdataframediagonaldoparallel
Avi
Avi

Reputation: 2283

Upper diagonal calculation in parallel in R

I use the following code:

  library(foreach)
  library(doParallel)
  N<-5
  cl<-makeCluster(8)
  registerDoParallel(cl)

  #loop
  s8 <- foreach(i=1:N, .combine='rbind') %:%
    foreach(j=1:N, .combine='c') %dopar% {

        dis <-as.numeric (i+j)  } ## In reality there something more complicated

  stopCluster(cl)

I get the following result:

> s8
         [,1] [,2] [,3] [,4] [,5]
result.1    2    3    4    5    6
result.2    3    4    5    6    7
result.3    4    5    6    7    8
result.4    5    6    7    8    9
result.5    6    7    8    9   10

I would like to get the upper diagonal matrix as followed:

> s8
         [,1] [,2] [,3] [,4] [,5]
result.1    2    3    4    5    6
result.2    0    4    5    6    7
result.3    0    0    6    7    8
result.4    0    0    0    8    9
result.5    0    0    0    0   10

If I change the iterator of the inner loop to 

foreach(j=i:N, .combine='c') %dopar% {

I don't get the required results.

Upvotes: 1

Views: 64

Answers (1)

Roland
Roland

Reputation: 132969

Maybe this (although I would probably only parallelize the outer loop):

library(foreach)
library(doParallel)
N<-5
cl<-makeCluster(4)
registerDoParallel(cl)

#loop
#result is a list of vectors
s8 <- foreach(i=1:N) %:%
  foreach(j=i:N, .combine='c') %dopar% {

    as.numeric (i+j)  } ## In reality there something more complicated

stopCluster(cl)

Post-processing to get same lengths:

s8 <- t(vapply(s8, function(x) {
  x <- rev(x)
  length(x) <- N
  rev(x)
}, FUN.VALUE = numeric(N)))

s8[lower.tri(s8)] <- 0

#     [,1] [,2] [,3] [,4] [,5]
#[1,]    2    3    4    5    6
#[2,]    0    4    5    6    7
#[3,]    0    0    6    7    8
#[4,]    0    0    0    8    9
#[5,]    0    0    0    0   10

Upvotes: 1

Related Questions