Can I get.seed() somehow?

In reference to the statement set.seed(), can I get the seed instead after running some code if I didn't set it explicitly?

I've been re-running some code (interactively / at the console) containing a function that randomises some sample of the input data (the function is part of the kohonen package). After playing with it for some time to see the variety of output (it was an 'unstable' problem), I noticed one result that was pretty interesting. I of course had not used set.seed(), but wondered if I could get the seed after running the code to reproduce the result?

In ?set.seed I see

.Random.seed saves the seed set for the uniform random-number generator

But I don't know how that helps.

Solution 1:

If you didn't keep the seed, there's no general way to "roll back" the random number generator to a previous state after you've observed a random draw. Going forward, what you may want to do is save the value of .Random.seed along with the results of your computations. Something like this.

x <- .Random.seed
result <- <your code goes here>
attr(result, "seed") <- x

Then you can reset the PRNG as follows; result2 should be the same as result.

.Random.seed <- attr(result, "seed")
result2 <- <your code goes here>

Solution 2:

Hong's answer above is robust. For quick and dirty solutions, where I just re-execute a whole script until I get interesting behavior, I randomly pick an integer, print it out, then use that as a seed. If my particular run has interesting behavior, I note that seed:

eff_seed <- sample(1:2^15, 1)
print(sprintf("Seed for session: %s", eff_seed))
set.seed(eff_seed)

Solution 3:

To add to the answer mpettis gave, if you don't want to re-execute the script manually--generating new random seeds each iteration--you could do something like this:

# generate vector of seeds
eff_seeds <- sample(1:2^15, runs)

# perform 'runs' number of executions of your code
for(i in 1:runs) {
    print(sprintf("Seed for this run: %s", eff_seeds[i]))
    set.seed(eff_seeds[i])

    # your code here
    # don't forget to save your outputs somehow
}

Where the variable 'runs' is a positive integer indicating the number of times you want to run your code.

This way you can generate a lot of output rapidly and have individual seeds for each iteration for reproducibility.

Solution 4:

> rnorm(5)
[1] -0.17220331 -0.31506128 -0.35264299  0.07259645 -0.15518961
> Seed<-.Random.seed
> rnorm(5)
[1] -0.64965000  0.04787513 -0.14967549  0.12026774 -0.10934254
> set.seed(1234)
> rnorm(5)
[1] -1.2070657  0.2774292  1.0844412 -2.3456977  0.4291247
> .Random.seed<-Seed
> rnorm(5)
[1] -0.64965000  0.04787513 -0.14967549  0.12026774 -0.10934254