CODEWITHSUNDEEP
rnested-loopslapplysapply
yungFanta
yungFanta

Reputation: 31

Using a nested loop in R alternative

I have the following code that in R that I'm trying to speed up since I know loops are slower in R. Is there a way to do this R without having to use nested loops. 

# initialize 2 vectors of length 10,000
totalNum <- rep(0,10000)
totalAmt <- rep(0,10000) 
values <- sample(200:5000,150000, replace = TRUE)
chances <-  # similar to f in length and contains values between 0 and 1 

# loop over length of a vector 
 for (i in 1:150000){
   # value between 200-5000 
   value <- values[i]
   # value of number between 0 and 1 
   chance <- chances[i]

  # loop over vectors created 
  for (j in 1:10000){
    # run test 
    dice <- runif(1)
    if (dice < chance){

    totalnum[j] <-  totalNum[j] + 1
    totalAmt[j] <- totalAmt[j] + value
   }

  }

}

I've been trying to use lapply or mapply but doesn't seem like it will work for this situation.

Upvotes: 0

Views: 1023

Answers (2)

Sathish
Sathish

Reputation: 12713

size = 150000 for values and chances vectors

library('data.table')
df1 <- data.table(totalNum = rep(0,10000),
                  totalAmt = rep(0,10000))
values <- sample(200:5000,150000, replace = TRUE)
chances <-  runif(n=150000, min=1e-12, max=.9999999999)

invisible( mapply( function(value, chance){
  df1[runif(10000) < chance, `:=` (totalNum = totalNum + 1, totalAmt = totalAmt + value)]
  return(0)
}, value = values, chance = chances) )

On my system, this code completes with the following time using the system.time() function.

# user  system elapsed 
# 252.83   43.93  298.68

Upvotes: 2

Dave2e
Dave2e

Reputation: 24079

lapply and mapply are just hidden loops with marginal improvement over a for loop. For significant improvement you need to use the vectorized forms of the functions.

The inner loop is easily replaced with a vectorized form:

#generate all of the rolls
dice<-runif(10000)

#identify the affected index
dicelesschance<-which(dice<chance)

#update the values 
totalNum[dicelesschance]<-totalNum[dicelesschance] + 1
totalAmt[dicelesschance]<-totalAmt[dicelesschance] + value

This should have a noticeable improvement on performance.

Upvotes: 1

Related Questions