Parameters of RNG through Data Analysis

[lucid dome]

Interesting homework. This was done in real life for eavesdropping bluetooth connections (I remember seeing a live demo in a security conference): https://www.rtl-sdr.com/spying-keyboard-presses-software-defined-radio/

as with any search process, you can start from a brute force search then see if you can reduce the search space somehow

[tough vortex]

Wouldn't Bayes' rule work well here? I would use a MAP estimator

[lucid dome]

That'd make sense to beat other humans. A RNG should be balanced. All the numbers should be statistically indistinguishable. Unless is a really poor RNG.

[tough vortex]

I thought by 'parameter' of RNG he meant a bias for either rock, paper or scissors

[lucid dome]

Oh

[tough vortex]

but yes then the goal would be to beat the other players

I mean that would be fun

[lucid dome]

That's not a RNG 😠 😛
but your reading makes more sense 😉

[tough vortex]

weighted RNG?

[lucid dome]

it's a prob. distr. Why call a prob. distr. a RNG?

you're sampling from the dist.

it's unusual in ML they would be so sloppy with the terminology

sampling from a distr. is super important in ML

nobody calls that a RNG

[tough vortex]

doesn't every RNG technically sample from a uniform distribution?

[lucid dome]

yup

so a biased sampler wouldn't call that to be called RNG

[coral dock]

from my understanding(not much) isn’t a weighted probability literally the definition of an rng

i’m not sure just trying to understand