Generation of random numbers in batches

Hello,

Imagine the following code in Matlab

rng(1)
a = rand(1,100)

rng(1)
a2 = rand(1,200);

These are two random generators initialized with the same seed. The elements in a are equal to the 100 first elements in a2.

Now, I want to create a random sequence a3 that satisfies a2 = [a, a3].

How should I choose the seed of a3 to satisfy a2? is this even possible?

Best

Answer 1

[NaN]

Found the answer.

Just need to store the state of the generator with s = rng and initialize with rng(s) the next generation

Thanks for your help

Answer 2

[Steven Lord]

The way to do what you're asking is to record the state of the random number generator. You can't really "work backwards".

>> rng(1)
>> a = rand(1, 100);
>> s = rng;

>> rng(1)
>> a2 = rand(1, 200);

>> rng(s) % Use the state recorded after generating the first 100
>> a3 = rand(1, 100);

>> isequal([a a3], a2)
ans =
  logical
   1

Answer 3

[Adam Danz]

The seeds below are chosen arbitrarily.

rng(1)
a = rand(1,100); 

rng(2)
a3 = rand(1,100); 

a2 = [a,a3]

You don't even have to change seeds if the a and a3 are generated consecutively.

rng(100)
a = rand(1,100); 
a3 = rand(1,100); 
a2 = [a,a3]; 

Answer 4

[Bjorn Gustavsson]

if you know from the start how many point in each set just do something like:

a2 = randn(1,num_points_in_a2);
a = a2(1:min(num_points_in_a,numel(a2)));
a3 = a2((1+num_points_in_a):end);

If you dont know how many point you want beforehand just create and throw away the excess points you need - if it is as few as a couple of 100 elements bothering about avoiding might take more time than will be lost...