Main Content

Control Random Number Streams on Workers

The random number generation functions rand, randi, and randn behave differently for parallel calculations compared to your MATLAB® client. You can change the behavior of random number generators on parallel workers or on the client to generate reproducible streams of random numbers.

By default, the MATLAB client and MATLAB workers use different random number generators, even if the workers are part of a local cluster on the same machine as the client. The table below summarizes the default settings for the client and workers:

 GeneratorSeedNormal Transform
Client'Twister' or 'mt19937ar'0'Ziggurat'
Worker (local or remote)'Threefry' or 'Threefry4x64_20'0'Inversion'

For more information about the available generators and normal transforms, see Choosing a Random Number Generator. Each worker in a cluster draws random numbers from an independent stream with the properties in the table. By default, the random numbers generated on each worker in a parfor loop are different from each other and from the random numbers generated on the client.

Note

If you have a GPU on your worker, different settings apply to random number streams on the GPU. For more information, see Random Number Streams on a GPU.

Client and Workers

If it is necessary to generate the same stream of numbers in the client and workers, you can set one to match the other. You can set the generator algorithm and seed using rng.

For example, you might run a script as a batch job on a worker, and need the same generator or sequence as the client. Suppose you start with a script file named randScript1.m that contains the line:

R = rand(1,4);

You can run this script in the client, and then as a batch job on a worker. Notice that the default generated random number sequences in the results are different.

randScript1; % In client
R
R =
    0.8147    0.9058    0.1270    0.9134
parallel.defaultClusterProfile('Processes')
c = parcluster();
j = batch(c,'randScript1'); % On worker
wait(j);load(j);
R
R =
    0.1349    0.6744    0.9301    0.5332

For identical results, you can set the client and worker to use the same generator and seed. Here, the file randScript2.m contains the following code:

rng(1,'Threefry');
R = rand(1,4);

Now, run the new script in the client and on a worker:

randScript2; % In client
R
R =
    0.1404    0.8197    0.1073    0.4131
j = batch(c,'randScript2'); % On worker
wait(j); load(j);
R
R =
    0.1404    0.8197    0.1073    0.4131

To reset the settings of the random number generator to their default values in the client and on the worker, you can add this code to the end of randScript2.m.

rng('default')

Different Workers

By default, each worker in a cluster working on the same job has an independent random number stream. If rand, randi, or randn are called in parallel, each worker produces a unique sequence of random numbers.

Note

Because rng('shuffle') seeds the random number generator based on the current time, do not use this command to set the random number stream on different workers if you want to ensure independent streams. This is especially true when the command is sent to multiple workers simultaneously, such as inside a parfor, spmd, or a communicating job. For independent streams on the workers, use the default behavior; or if that is not sufficient for your needs, consider using a unique substream on each worker using RandStream.

This example uses two workers in a parallel pool to show they generate unique random number sequences.

p = parpool(2);
spmd
    R = rand(1,4); % Different on each worker
end
R{1},R{2}
ans = 
    0.1349    0.6744    0.9301    0.5332
ans =
    0.6383    0.5195    0.1398    0.6509
delete(p)

If you need all workers to generate the same sequence of numbers, you can set each worker to use the same generator settings:

p = parpool(2);
spmd
    rng(0,'Philox'); % Default seed 0.
    R = rand(1,4); % Same on all workers
end
R{1},R{2}
ans =
    0.3655    0.6975    0.1789    0.4549
ans =
    0.3655    0.6975    0.1789    0.4549
delete(p)

If you need to control the random numbers at each iteration of a parfor-loop, see Repeat Random Numbers in parfor-Loops.

Normally Distributed Random Numbers

If you are working with normally distributed random numbers using the randn function, you can use the same methods as above using RandStream to set the generator type, seed, and normal transformation algorithm on each worker and the client.

For example, suppose the file randScript3.m contains the code:

stream = RandStream('Threefry','Seed',0,'NormalTransform','Inversion');
RandStream.setGlobalStream(stream);
R = randn(1,7)
You can run this code on the client and on a worker in a parallel job (using batch or spmd) to produce the same sequence of random numbers:
R =
    -0.3479    0.1057    0.3969    0.6544   -1.8228    0.9587    0.5360

See Also

|

Related Topics