Reproducible results in parallel

Now you are ready to make your results reproducible. You will use the simple embarrassingly parallel application for computing a mean of random numbers, mean_of_rnorm(), which we parallelized in the second chapter using clusterApply(). The parallel package, mean_of_rnorm(), n_vec (vector of sample sizes of 1000 replicated 5 times) are available in your workspace. You will now call clusterApply() repeatedly to check if results can be reproduced, without and with initializing the RNG.

1
- Run three replicates of this random number generation expression.
- In each replicate, use clusterApply() to call mean_of_rnorm() on n_vec, in parallel.
- Note that the results are transposed (using the t function), so that rows correspond to the 3 replicates and columns correspond to the 5 tasks defined by the length of n_vec.
2
- Update the replicated expression so that you set the cluster's random number generation stream's seed to 1234 before you call clusterApply().
- Look at each row of the output. You should start to experience a sensation of deja vu.

Exercise

Reproducible results in parallel

Instructions 1/2