mapehe / evt-ica-simu

Triton simu code

This repo was used in executing the simulations for a paper related to extreme value theory and independent component analysis. More generally, it illustrates one way to use a SLURM cluster (such as Aalto University's Triton). The simulation required generating samples independently and evaluating certain statistical quantities from them; the different scenarios were executed in parallel. A setup that solves any "embarrassingly parallel" problem can be built in a similar way and this repo can be useful starting point.

The approach here is simple: Commands of the form

Rscript [arguments]

Were executed, where [arguments] specifies the scenario to simulate. The following files were used for this:

• simu_round.r contains the simulation code. Additionally it illustrates how to process command line arguments in R.

• generate_args.py generates the command line arguments that are passed to simu_round.r

• submit.sbatch submits the job to SLURM and starts the execution

Running the simulation

Run

python generate_args.py

This will generate the command line arguments for the simulation. After the script has been executed, they have appeared in the folder args. To execute the simulation run

sbatch submit.sbatch

The results will appear in folder out.