webppl/src/dists/gamma.js

Probabilistic programming for the web

'use strict';
var ad = require('../ad');
var base = require('./base');
var types = require('../types');
var util = require('../util');
var gaussian = require('./gaussian');
function sample(shape, scale) {
    if (shape < 1) {
        var r;
        r = sample(1 + shape, scale) * Math.pow(util.random(), 1 / shape);
        if (r === 0) {
            util.warn('gamma sample underflow, rounded to nearest representable support value');
            return Number.MIN_VALUE;
        }
        return r;
    }
    var x, v, u;
    var d = shape - 1 / 3;
    var c = 1 / Math.sqrt(9 * d);
    while (true) {
        do {
            x = gaussian.sample(0, 1);
            v = 1 + c * x;
        } while (v <= 0);
        v = v * v * v;
        u = util.random();
        if (u < 1 - 0.331 * x * x * x * x || Math.log(u) < 0.5 * x * x + d * (1 - v + Math.log(v))) {
            return scale * d * v;
        }
    }
}
var Gamma = base.makeDistributionType({
    name: 'Gamma',
    desc: 'Distribution over positive reals.',
    params: [
        {
            name: 'shape',
            desc: '',
            type: types.positiveReal
        },
        {
            name: 'scale',
            desc: '',
            type: types.positiveReal
        }
    ],
    wikipedia: true,
    mixins: [base.continuousSupport],
    sample: function () {
        return sample(ad.value(this.params.shape), ad.value(this.params.scale));
    },
    score: function (x) {
        var shape = this.params.shape;
        var scale = this.params.scale;
        return ad.scalar.sub(ad.scalar.sub(ad.scalar.sub(ad.scalar.mul(ad.scalar.sub(shape, 1), ad.scalar.log(x)), ad.scalar.div(x, scale)), ad.scalar.logGamma(shape)), ad.scalar.mul(shape, ad.scalar.log(scale)));
    },
    support: function () {
        return {
            lower: 0,
            upper: Infinity
        };
    }
});
module.exports = {
    Gamma: Gamma,
    sample: sample
};