Random with a high probabilty number(s)?

Question

Possible Duplicate:
Biased Random Number Generator

for once I need a random number generator with at least two numbers that have higher probabality than the others.

i.e for example: random 1->10 in a sequence of 1000. Numbers A=3 and B=7.

A - should repeat approx. at least 20% of the time. B - should repeat approx. at least 30% of the time.

That should cover at least 50% of the 1000 sequence. Also the insertion of A and B should be somewhat probable/random in themselves. Not just add A and B every Nth step. Total/exact control isn't needed.

Any Ideas?

I'm a noob - c++ style code would be greatly appreciated!

score 2 · Accepted Answer · answered Jul 04 '12 at 22:49

One way you can do this is to randomly generate a number between 0.0 and 1.0, and choose what number to generate based on that number. For example, to implement your example scenario (pseudocode):

let "result" be an array of 1000 integers
let "i" be an integer
for i = 1 to 1000:
    let "j" be a random number between 0.0 and 1.0
    if j < 0.2:
        let result[i] be 3
    else if j < 0.5:
        let result[i] be 7
    else:
        let "k" be an integer
        do, while k = 3 or k = 7:
            let "k" be a random number in the range 1 to 10
        let result[i] be k
end

Basically, j is used to partition the range 1 to 10 into three parts - one part covering from 0% to 20% of the range (the first if), the second covering from 20% to 50% of the range (i.e. 30% wide, the second if), and the last covering the remaining 50%. Depending on which part we randomly fall in, we chose the number to generate as appropriate.

That still depends on the original random generator. Also forgot to be MORE specific. It's use is for an Arpregiator - used in music synthesizers. Say the values are array[24] (that's two octaves in music, a distubution mapping would bunch the values too close to the given points - A & B in my case. A & B should have probability(interger) and the rest should be random(integer). — ichad.c, Jul 04 '12 at 23:01
Sorry, just read the update to your reply - that should work. Thanks! — ichad.c, Jul 04 '12 at 23:06

bames53 · Answer 2 · 2016-01-21T18:29:09.107

You should use the <random> library for this.

#include <random>

#include <iostream>
#include <algorithm>
#include <iterator>

int main() {
    // create a discrete distribution where the third object has 20% probability and
    //   the seventh has 30%
    std::vector<double> probabilities(10, 5.0/8.0);
    probabilities[2] = 2.0;
    probabilities[6] = 3.0;
    std::discrete_distribution<int> dist(begin(probabilities),end(probabilities));

    // our underlying source of randomness
    std::random_device r;
    std::seed_seq seed{r(), r(), r(), r(), r(), r(), r(), r()};
    std::mt19937 eng(seed);

    // create a function object that uses the distribution and source of randomness to
    //   produce values from 1 to 10
    auto rand = [&]{ return dist(eng) + 1; };

    std::vector<int> x;

    // store 1000 random values
    for (int i=0;i<1000;++i)
        x.push_back(rand());

    // count how many of each value, to verify that 3 comes out ~200 times and 7 comes
    //   out ~300 times
    for (int i=1;i<=10;++i)
        std::cout << i << ": " << count(begin(x),end(x),i) << '\n';

    // print all the values
    copy(begin(x),end(x),std::ostream_iterator<int>(std::cout, " "));
}

Random with a high probabilty number(s)?

2 Answers2