3 ms·
It's an important question, but not too crazy for me. Think about Wolfram's automata.. 1 dimensional, recurrent, 3-wide inputs (ie super simple).. and one of t
by pmayrgundter 4y ago
It's an important question, but not too crazy for me. Think about Wolfram's automata.. 1 dimensional, recurrent, 3-wide inputs (ie super simple).. and one of the 256 possible rules generates randomness that is used as the PRNG in Mathematica.
Depends on the details of the ANN, but if it's a recurrent net then seems like there's at least one very shallow net that solves it.
- kevin_thibedeau 4y agoPure rule 30 won't pass BigCrush.
- pmayrgundter 4y agoYeah, if you mean including all output bits, agreed, but note it's not used that way and there's a well-known area of reliable randomness to the right end of the output bits that is used. To the original point, I think a "right pass filter" is simple for an ANN to figure out. Some interesting discussion on Rule30 and BigCrush, updated as of 2017 by Daniel Lichtblau at Wolfram Research: "In house testing has shown ExtendedCA passing all Diehard and BigCrush tests... The Rule 30 generator is almost as good as this in terms of the Big Crush suite... The upshot is that the Rule 30 and ExtendedCA RNGs are, by the standards of current testing, quite sound." https://mathematica.stackexchange.com/questions/3208/quality-of-random-numbers https://mathematica.stackexchange.com/questions/3208/quality... ExtendedCA seems like also a good demonstration of a very simple high quality PRNG that an RNN could discover: https://reference.wolfram.com/language/tutorial/RandomNumberGeneration.html#18361715 https://reference.wolfram.com/language/tutorial/RandomNumber... It's 1d, 5-wide input, inputs 4 cells apart from each other
- lgvld 4y agoYes, you’re right about it, thank you. I read true RNG while the paper is talking about PRNG. And a RNN would fit the bill, indeed.