Randomization measure in rand

What is the randomization power in rand function.? For example UNIX drand48 can generate a randomization of 2^48 (~10^14).

Answer 1

[Walter Roberson]

2^53 different possibilities.

You should also be caring about how long the period is (that is, the time until the sequence repeats.) It is on the order of 2^11000 for the default randomization routine.

[Sandeep Ganji]

Thank you for the quick reply. Does using the rng('shuffle') improves the randomness?

[Shashank Prasanna]

rng('shuffle') just resets the seed of the RNG based on the current time on your machine. Doesn't do anything to the algorithm itself.

[Walter Roberson]

If your thought is to periodically have your program use rng('shuffle') in order to "improve the randomness", then interestingly it can be shown that this will reduce the randomness, because the times used as input to the shuffling will end up being correlated.

If you need "cryptographic strength" randomness, then you need to use special routines for that purpose that are "hardened" to reduce leakage of information about the internal state of the generator.

[Sandeep Ganji]

Thank you. I will consider shifting to a more robust random number generator.