Mersenne twister random numbers not the same as C++

Hi,

Trying to get the same random numbers in Matlab and C++ by using the Mersenne twister, but Matlab seems to skip every second output of Mersenne twister and also small differences are seen in the 32 bit words that (almost) match the C++ output, see example below.

Is there some way to get the same numbers in Matlab as you get when using the C++ standard?

% Print first 10 numbers from Mersenne twister in interval [0, 2^32-1] with 7 as seed

rng(7, 'twister');

randi([0 4294967295],1,10, 'uint32')'

% Matlab 9.3.0.713579 (R2017b) printout:

% 327741607

% 3349725706

% 1882953311

% 3107259278

% 4200432969

% 2312822152

% 2152296002

% 309457261

% 1152936640

% 2146978992

% C++ implementation and printout

%#include <iostream>

%#include <random>

% using namespace std;

% int main()

% {

% // % Print first 10 numbers from Mersenne twister in interval [0, 2^32-1] with 7 as seed

% std::mt19937 mt_rand(7);

% for (int n = 0; n < 10; n++)

% {

% cout << mt_rand() << endl;

% }

% return 0;

% }

% 327741615

% 976413892

% 3349725721

% 1369975286

% 1882953283

% 4201435347

% 3107259287

% 1956722279

% 4200432988

% 1322904761

Answer 1

[Shantanu Dixit]

Hi Tobias, I’ve encountered a similar scenario where the random numbers generated by MATLAB and C++ don’t match. This might be due to MATLAB’s default Mersenne Twister using an implementation that prioritizes full precision floating-point numbers.

As a workaround, I was able to align MATLAB’s output with C++ by configuring the random stream explicitly:

R = RandStream('mt19937ar', 'Seed', 7);
R.FullPrecision = false;               % Set to not use full precision
RandStream.setGlobalStream(R);         % Set this as the global stream
randi([0 4294967295], 1, 10, 'uint32')'

ans = 10×1

327741615 976413892 3349725721 1369975286 1882953283 4201435347 3107259287 1956722279 4200432988 1322904761

<mw-icon class=""></mw-icon>

<mw-icon class=""></mw-icon>

You can refer to the following documentation to know more about setting the global stream for 'randStream':

'randstream': https://www.mathworks.com/help/matlab/ref/randstream.html

'setGlobalStream': https://www.mathworks.com/help/matlab/ref/randstream.randstream.setglobalstream.html

Hope this helps!