random number generation initial state

2 views (last 30 days)
mehmet umut caglar
mehmet umut caglar on 12 Sep 2012
Answered: Chuck on 5 May 2016
what is the possibility of rng('shuffle') to set the system to a specific state.
say I have a code
rng('shuffle')
A=random('unif',0,1,1,5);
say I run this code in 2 different computers. What is the possibility for these 2 guys give the same A?

Answers (2)

Matt Tearle
Matt Tearle on 12 Sep 2012
Digging through the code, rng(shuffle) calls RandStream.shuffleSeed. In there you can find a comment:
% Create a seed based on 1/100ths of a second, this repeats itself
% about every 497 days.
So, if we believe that, the chances of getting the same seed are about 1 in 3600*24*497*100 = 4.3 billion. Now that's assuming you're just running these at two randomly chosen times (on the same computer or not).
If you run the code on one computer, then go to another and run it there, that comment implies that there's no chance they'll be the same (unless you can run the two programs within 0.01 seconds of each other).
(Of course, the two computers' clocks are probably not perfectly in sync, either.)
  1 Comment
Matt Fig
Matt Fig on 12 Sep 2012
Edited: Matt Fig on 12 Sep 2012
My only question was whether or not Random uses RandStream or something else because Random is a Simulink function. I have very little experience with SimuLink so I don't know how it interacts with the MATLAB core.
EDIT I see there is a non-Simulink function with the same name in the Stats TB. This is probably what is referred to here.

Sign in to comment.


Chuck
Chuck on 5 May 2016
It is extremely unlikely if you are using rng("shuffle"). Practically, you should not get the exact same seed.

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!