random

hey /prog/.

is there a random number generator algorithm/library that supports something like this: use this seed, give me next nth random number? obviously it shouldn't iterate n times. I don't care if it is not that random.

there is probably something very simple for this but I can't get my head around it.

So you're looking for a specific random number generator

int Random(){
    return 4; //Gauranteed to be random
}

>>1

would this work?


unsigned int random(unsigned int n) {
  return n*__REALLY_BIG_NUMBER_THAT_IS_RELATIVELY_PRIME_TO_MAX_INT__;
}

>>3
a little more random please
>>4
I don't think so

So in the end I will use the seed iterate through random values n times.

Also is there a way to get current seed of rand() in C? So if I use srand(seed), the random sequence will start from that point.

There's an LCG variant with that property: https://en.wikipedia.org/wiki/Lehmer_RNG

I haven't checked my math, but I think you want: X_k+i = g⁽ⁱ⁺¹⁾·X_k mod n

>>5

how about?


static unsigned int seed;

void srand(unsigned int new_seed) {
  seed = new_seed;
}

unsigned int random(unsigned int n) {
  return (n+seed)*__REALLY_BIG_NUMBER_THAT_IS_RELATIVELY_PRIME_TO_MAX_INT__;
}

>>7
>>5
hmm, interesting. I thought you were trolling. >>6 I guess this is legit. I will try it, thanks.

>>7
Keep going bud, eventually you'll reinvent the LCG.

>>8

yeah it's a really simply technique. It makes results that look kind of random, and it visits every number exactly once, until it starts to repeat. Like if MAX_INT is 12, and you let the large number be 7, then the sequence is:


0,
7,
14-12=2,
9,
16-12=4,
11,
18-12=6,
13-12=1,
8,
15-12=3,
10,
17-12=5,
12-12=0,

0,7,2,9,4,11,6,1,8,3,10,5

It looks sorta random. cept for

0,7,2, 9
1,8,3,10

looks a little suspicious.


>>9 thanks for the reference. I must better my primitive techniques.

>>10
Yeah, you're headed in the right direction and it's a good exercise. You just need to replace MAX_INT with any modulus to generalize it (and to prevent the modulus from being a power of 2), reorder things a bit and you're basically there.

0,7,2, 9
 1,8,3,10

looks a little suspicious.

That can be helped by choosing the right constants though it will always have flaws in the distribution but it's still pretty good.

　 ∧＿∧　　　    ∧＿∧　　　∧＿∧　　 ∧＿∧　 　 　     ∧＿∧
　（　・∀・） 　    （　｀ー´）　　（　´∀｀）　 （　ﾟ ∀ﾟ ）　　　     （　＾∀＾）
　（　　　　つ┳━━∪━━∪━∪━━∪━∪━∪━━━━━┳⊂　 　 　つ
　｜ ｜　| ┃This thread has peacefully ended.┃　｜ ｜ ｜
　（_＿）＿）┻━━━━━━━━━━━━━━━━━━━━┻　（_＿）＿）　　　　　Thank you.

cat /dev/random

>>3
Hello, randall!!!

>>14
Bitches don't know about my rubber ducky stacking

Dammit. The old thread was a lot more interesting.

C#/.NET's Random class does exactly that, OP.

>>1
What for do you need that?

Any old linear congruential generator can do that, obviously: represent one step as a matrix over the modulo field, rise the matrix to n-th power in log(n) steps. But LGCs suck.

The state of the art in PRNG is Mersenne Twister, it can't be fast-forwarded like that apparently, but given its period you can just start with consecutive natural numbers as seeds and be sure to get non-intersecting subsequences of reasonable length. If you want to get some deterministic non-intersecting subsequences. What do you want, OP?

>>17 is a .NET and Microsoft fag, be warned.

>>19
Thanks man I was gonna hire him!!!!!!!!!!!!!!!!!!

Set ⎕RL to the value you want and then execute ? the desired number of times.

>not using chaotic equations
>2012

>>18
He probably wanted the O(1) solution that meets his needs, which was posted already.