multivariate-normal/distribution.js

Port of NumPy's random.multivariate_normal to Node.JS

"use strict";

Object.defineProperty(exports, "__esModule", {
    value: true
});

var _validation = require("./validation.js");

var _gaussian = require("gaussian");

var _gaussian2 = _interopRequireDefault(_gaussian);

var _numeric = require("numeric");

var _numeric2 = _interopRequireDefault(_numeric);

function _interopRequireDefault(obj) { return obj && obj.__esModule ? obj : { default: obj }; }

// Low-level distribution constructor. NOT a public API.
//
// n: dimensionality of the distribtion
// mean: vector of mean, of length n.
// cov: covarance matrix, of size n-by-n.
// svd: { u, s, v } from decomposition of cov
//
// preconditions:
//   - mean and cov have been validated
//   - mean and cov are frozen

var standardNormalDist = (0, _gaussian2.default)(0, 1);

var standardNormalVector = function standardNormalVector(length) {
    var ary = [];

    for (var i = 0; i < length; i++) {
        ary.push(standardNormalDist.ppf(Math.random()));
    }

    return ary;
};

var Distribution = function Distribution(n, mean, cov, _ref) {
    var u = _ref.u,
        s = _ref.s,
        v = _ref.v;

    return {
        sample: function sample() {
            // From numpy (paraphrased):
            //   x = standard_normal(n)
            //   x = np.dot(x, np.sqrt(s)[:, None] * v)
            //   x += mean
            //
            // https://github.com/numpy/numpy/blob/a835270d718d299535606d7104fd86d9b2aa68a6/numpy/random/mtrand/mtrand.pyx

            // np.sqrt(s)[:, None] * v
            //
            // This is an elegant way in numpy to multiply each column of
            // v by sqrt(s). Unfortunately, we don't have numpy, so we do this
            // manually
            var sqrtS = s.map(Math.sqrt);
            var scaledV = v.map(function (row) {
                return row.map(function (val, colIdx) {
                    return val * sqrtS[colIdx];
                });
            });

            // We populate a row vector with a standard normal distribution
            // (mean 0, variance 1), and then multiply it with scaledV
            var standardNormal = standardNormalVector(n);

            // compute the correlated dsitribution based on the covariance
            // matrix
            var variants = _numeric2.default.dot(standardNormal, _numeric2.default.transpose(scaledV));

            // add the mean
            return variants.map(function (variant, idx) {
                return variant + mean[idx];
            });
        },
        getMean: function getMean() {
            return mean;
        },
        setMean: function setMean(unvalidatedMean) {
            var newMean = (0, _validation.validateMean)(unvalidatedMean, n);
            return Distribution(n, newMean, cov, { u: u, s: s, v: v });
        },
        getCov: function getCov() {
            return cov;
        },
        setCov: function setCov(unvalidatedCov) {
            var _validateCovAndGetSVD = (0, _validation.validateCovAndGetSVD)(unvalidatedCov, n),
                newCov = _validateCovAndGetSVD.cov,
                newSVD = _validateCovAndGetSVD.svd;

            return Distribution(n, mean, newCov, newSVD);
        }
    };
};

exports.default = Distribution;