Different behaviour of rng() and rand() in parfor-loop

Hey folks,

I experienced a different output of the standard function rand() in a parfor-loop although I am using the same seed-value. Example:

>> parfor i = 1:1, rng(123), rand(1,1), end

ans =
    0.2751

>> for i = 1:1, rng(123), rand(1,1), end

ans =
    0.6965

Is this a bug? I would expect exactly the same output of rand() in parfor.

Thanks & greetings Jonas

------------------------------------------------

ps: here are my detailed version information, I use an Intel i7-2600 processor.

----------------------------------------------------------------------------------------------------

MATLAB Version: 8.2.0.701 (R2013b)

MATLAB License Number: ••••••

Operating System: Microsoft Windows 7 Version 6.1 (Build 7601: Service Pack 1)

Java Version: Java 1.7.0_11-b21 with Oracle Corporation Java HotSpot™ 64-Bit Server VM mixed mode

----------------------------------------------------------------------------------------------------

MATLAB Version 8.2 (R2013b)

DIPUM Toolbox Version 1.1.4

DIPUM Toolbox Version 1.1.3

Image Processing Toolbox Version 8.3 (R2013b)

Optimization Toolbox Version 6.4 (R2013b)

Parallel Computing Toolbox Version 6.3 (R2013b)

[John Fox]

I had the exact same problem. My for loops gave a different answer than my parfor loops. The reason is

http://www.mathworks.com/help/distcomp/repeat-random-numbers-in-parfor-loops.html

As described in Control Random Number Streams, each worker in a cluster has an independent random number generator stream. By default, therefore, each worker in a pool, and each iteration in a parfor-loop has a unique, independent set of random numbers. Subsequent runs of the parfor-loop generate different numbers.

 I fixed this with rng(123,'twister'). At least this worked for me.

[John Fox]

Notice what happens when you add 'twister' to the parfor loop

>> for i = 1:1, rng(123), rand(1,1), end

ans =

    0.6965

Which is the same answer you got.

>> parfor i = 1:1, rng(123,'twister'), rand(1,1), end

ans =

    0.6965

It is the same answer when you add 'twister' to the parfor loop

Answer 1

[Edric Ellis]

The client and the workers are set up to use different random generators, so this is expected. This is covered in the documentation. Note that "rng(seed)" changes only the seed but not the underlying generator type.

[Walter Roberson]

https://www.mathworks.com/help/parallel-computing/control-random-number-streams-on-workers.html is the current link.