Generating Probability Distributions in Java

Question

I want to find out how to generate a probability distribution. I am working on the Stochastic Shortest Path Problem where edges have associated probability distributions with associated costs for each probability. I was able to generate a (normal) distribution like this:

0     1     2     3     4
0.15  0.2   0.18  0.22  0.25

such that the sum of all probabilities is 1 by following answers provided in this question. Now, I need to generate other distributions like bi-normal, log-normal and gamma. I would really appreciate any clarifications on these distributions and (pseudo) code, preferably in Java, on generating them.

Robert Altena · Answer 1 · 2015-11-17T12:51:51.550

Normal distribution: http://commons.apache.org/proper/commons-math/apidocs/org/apache/commons/math3/distribution/NormalDistribution.html

Log normal distribution: http://commons.apache.org/proper/commons-math/apidocs/org/apache/commons/math3/distribution/LogNormalDistribution.html

Gamma distribution: http://commons.apache.org/proper/commons-math/apidocs/org/apache/commons/math3/distribution/GammaDistribution.html

I would stay away from 100% home-cooked solutions such as the question you refer to.

Here is some code to generate the distributions. You will have to normalize them yourself to make the sum add up to 1.

import org.apache.commons.math3.distribution.AbstractRealDistribution;
import org.apache.commons.math3.distribution.NormalDistribution;
import org.apache.commons.math3.distribution.LogNormalDistribution;
import org.apache.commons.math3.distribution.GammaDistribution;

public class DistributionGenerator {

    public static void main(String[] args) {

    // Pick one and comment out the others:
    //AbstractRealDistribution distr = new NormalDistribution(2.0,0.5);  //  mean and standard deviation constructor.
    //AbstractRealDistribution distr = new LogNormalDistribution(0.0,1.0); // scale and shape constructor.
    AbstractRealDistribution distr = new GammaDistribution(2.0, 1.0);

    for(int i=0; i<5; ++i)
    {
        System.out.println( distr.density(i));
    }
    }
    }

Thanks for your help! However, I still can't figure out how to get an array of size _n_ of random numbers whose sum is 1 with the above libraries. Could you clarify that, please? — ayePete, Nov 17 '15 at 08:47

score 2 · Answer 2 · answered Nov 16 '15 at 07:44

2

Quick search reveals the Colt library and its suite of distributions classes.

answered Nov 16 '15 at 07:44

wilx

17,697
6
59
114

Generating Probability Distributions in Java

2 Answers2