Mersenne Twister RNG init_by_array

Hi,

Can I init Mersenne Twister with array as a seed? (The C mt19937ar implementation support it: http://www.math.sci.hiroshima-u.ac.jp/~m-mat/MT/MT2002/emt19937ar.html)

Thanks

Answer 1

[Jan]

No, this is not possible in Matlab. As far as I know, you can seed it with 32 bits only - but I'm still such surprised about this, that I hope I've overseen a detail.

See https://mathworks.com/matlabcentral/answers/329848-how-to-get-non-repeatable-randn-results

[Noam Mazor]

Thanks! Is there a simple way to use the C implementation as an external lib and to initialize it from there?

Answer 2

[Peter Perkins]

The "array" you are talking about is typically a state vector, right? You can set the generator state:

>> r = RandStream.getGlobalStream
r = 
mt19937ar random stream (current global stream)
             Seed: 0
  NormalTransform: Ziggurat
>> s = r.State;
>> rand(1,5)
ans =
      0.09754       0.2785      0.54688      0.95751      0.96489
>> r.State = s;
>> rand(1,5)
ans =
      0.09754       0.2785      0.54688      0.95751      0.96489

[Noam Mazor]

Hi, No, the C implementation has a function (init_by_array) that maps arbitrary array (in any length) into a state vector.

Thanks!