Create statistically independent random number streams
creates a single random stream that uses the uniform pseudorandom number generator algorithm
s = RandStream.create(
RandStream.list returns all possible values for
see Creating and Controlling a Random Number Stream for details on generator algorithms. The
RandStream function is a more concise alternative when you need to create a
s1,s2,...] = RandStream.create(
n random number streams. The streams are independent in a
pseudorandom sense. The streams are not necessarily independent from streams created at
Not all generator types support multiple streams. Use either the multiplicative
lagged Fibonacci generator (
'mlfg6331_64') or the combined multiple
recursive generator (
'mrg32k3a') to create multiple streams.
Create three independent streams. Generate random numbers from each stream. Check the correlations between them. The correlations between different streams are not exactly 0 because they are calculated from a sampling of the distribution.
[s1,s2,s3] = RandStream.create('mrg32k3a','NumStreams',3); r1 = rand(s1,100000,1); r2 = rand(s2,100000,1); r3 = rand(s3,100000,1); corrcoef([r1,r2,r3])
ans = 3×3 1.0000 0.0016 -0.0019 0.0016 1.0000 -0.0012 -0.0019 -0.0012 1.0000
You can also create one stream from three independent streams and designate it as the global stream.
s2 = RandStream.create('mrg32k3a','NumStreams',3,'StreamIndices',2); RandStream.setGlobalStream(s2);
Generate random numbers from the global stream.
r = rand(100000,1);
Create three mutually independent streams to simulate one-dimensional random walks using the method
[s1,s2,s3] = RandStream.create('mrg32k3a','NumStreams',3);
Generate a random walk from the first stream. First, set it as the global stream. Use the first stream to generate 5,000 random steps from the standard normal distribution. Choose the starting position at 0 and use
cumsum to calculate the cumulative sum of the random steps. Plot the resulting random walk.
RandStream.setGlobalStream(s1) dy1 = randn(5000,1); y1 = cumsum([0; dy1]); plot(y1)
Repeat the process using the second and third streams. Plot the results on the same axes.
hold on; dy2 = randn(s2,5000,1); y2 = cumsum([0; dy2]); plot(y2) dy3 = randn(s3,5000,1); y3 = cumsum([0; dy3]); plot(y3) hold off
Calculate the correlations among the streams. The correlations between different streams are not exactly 0 because they are calculated from a sampling of the distribution.
C = corrcoef([dy1 dy2 dy3])
C = 3×3 1.0000 -0.0363 0.0155 -0.0363 1.0000 -0.0012 0.0155 -0.0012 1.0000
gentype— Random number generator algorithm
Random number generator algorithm, specified as a character vector or string scalar naming a random number generator. MATLAB® offers several generator algorithms. The following table summarizes the names and key properties of the available generator algorithms. For more information, see Creating and Controlling a Random Number Stream.
|Name||Generator||Multiple Stream and Substream Support||Approximate Period In Full Precision|
|Mersenne Twister (used by default stream at MATLAB startup)||No||219937-1|
|SIMD-Oriented Fast Mersenne Twister||No||219937-1|
|Multiplicative lagged Fibonacci generator||Yes||2124 (251 streams of length 272)|
|Combined multiple recursive generator||Yes||2191 (263 streams of length 2127)|
|Philox 4x32 generator with 10 rounds||Yes||2193 (264 streams of length 2129)|
|Threefry 4x64 generator with 20 rounds||Yes||2514 (2256 streams of length 2258)|
|Shift-register generator summed with linear congruential generator||No||264|
|Modified subtract with borrow generator||No||21492|
|Multiplicative congruential generator||No||231-2|
comma-separated pairs of
the argument name and
Value is the corresponding value.
Name must appear inside quotes. You can specify several name and value
pair arguments in any order as
'NumStreams'— Number of streams
1(default) | positive integer
Number of independent streams to create, specified as the comma-separated pair
'NumStreams' and a positive integer.
'StreamIndices'— Stream indices
[1:N](default) | vector of positive integers | positive integer
Stream indices, specified as the comma-separated pair consisting of
'StreamIndices' and a vector of positive integers or a positive
integer. Specify this parameter to index the current stream from among the group of
streams with which it was created. The default value is
N is the value of
'Seed'— Random number seed
0(default) | nonnegative integer |
Random number seed, specified as the comma-separated pair consisting of
'Seed' and a nonnegative integer or
The seed specifies the starting point for the algorithm to generate random numbers.
'shuffle' creates a seed based on the current time. If you
specify an integer, it must be between 0 and
232 − 1.
Specify the generator seed as an initialization step when creating a stream at
MATLAB startup or before running a simulation. To reproduce a stream, use the same
seed every time. While using multiple seeds will create multiple sequences of random
numbers, there is no guarantee that the different sequences are statistically
independent. In situations where this is important, use
RandStream.create with multiple outputs to create multiple
streams that are statistically independent.
'NormalTransform'— Normal transformation algorithm
Transformation algorithm to generate normally distributed random numbers from the
random number stream using
randn, specified as the comma-separated
pair consisting of
'NormalTransform' and one of the algorithm names
'CellOutput'— Option to return cell array
Option to return cell array, specified as the comma-separated pair
'CellOutput' and logical
1). If you
RandStream.create returns the stream objects as elements of a
Typically, you call
RandStream.create once to create multiple
independent streams in a single pass or at the beginning of a MATLAB session. For example, you can create three independent streams by using
[s1,s2,s3] = RandStream.create('mrg32k3a','NumStreams',3).
Alternatively, you can create each stream from a separate call to
RandStream.create, but you must specify the appropriate values for
'StreamIndices' to ensure their independence:
Specify the same values for
'Seed' in each case.
Specify a different value for
'StreamIndices' each time. All
values should be between
1 and the value of
For example, create two independent streams by using