Is it possible to generate specific random numbers?

The generation of unique random numbers is done simply through randperm.

example

function idx=RandNumberSequence(Key,N)

% using Key as seed to the used random number generator.

Key = max(cumsum(double(Key))) ;

s=rng(Key); % Set the random number generator seed

idx=randperm(N);

end

The question is, given certain N random numbers, is it possible to select the Key that generates these N numbers?

let N=10, let te random numbers be 89, 70, 10, 12, 67, 25, 200, 18, 90, 86

Key=? (to be selected)

Answer 1

[Steven Lord]

Sure.

seedvalue = 0;
expected = [89 70 10 12 67 25 200 18 90 86];
rng(seedvalue, 'twister');
actual = randperm(max(expected), numel(expected));
while ~isequal(actual, expected)
    seedvalue = seedvalue + 1;
    rng(seedvalue, 'twister');
    actual = randperm(max(expected), numel(expected));
end

It may take a while. This assumes you haven't manually created a state vector (which would be an advanced maneuver.)

If you're asking if there's a clever way to directly go from the sequence to the seed value, I'm not aware of any offhand.

If you know you're going to need or want to reproduce a given sequence, record the seed before generating the sequence.

[Omar Ali Muhammed]

However, Speed is the main challenge espicially for long sequence. Is it possible to parallelize the code?

[Steven Lord]

Yes, this code could be parallelized through the use of parfor to sweep the seed space. But if speed is of the essence the best solution is to record the seed before you generate the sequence rather than trying to determine the seed after the fact.