Skip to content
This repository

HTTPS clone URL

Subversion checkout URL

You can clone with HTTPS or Subversion.

Download ZIP
tree: b7a5835258
Fetching contributors…

Cannot retrieve contributors at this time

file 240 lines (194 sloc) 9.01 kb
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240
using System;

namespace Moserware.Numerics
{
    public class GaussianDistribution
    {
        // Intentionally, we're not going to derive related things, but set them all at once
        // to get around some NaN issues

        private GaussianDistribution()
        {
        }

        public GaussianDistribution(double mean, double standardDeviation)
        {
            Mean = mean;
            StandardDeviation = standardDeviation;
            Variance = Square(StandardDeviation);
            Precision = 1.0/Variance;
            PrecisionMean = Precision*Mean;
        }

        public double Mean { get; private set; }
        public double StandardDeviation { get; private set; }

        // Precision and PrecisionMean are used because they make multiplying and dividing simpler
        // (the the accompanying math paper for more details)

        public double Precision { get; private set; }

        public double PrecisionMean { get; private set; }

        private double Variance { get; set; }

        public double NormalizationConstant
        {
            get
            {
                // Great derivation of this is at http://www.astro.psu.edu/~mce/A451_2/A451/downloads/notes0.pdf
                return 1.0/(Math.Sqrt(2*Math.PI)*StandardDeviation);
            }
        }

        public GaussianDistribution Clone()
        {
            var result = new GaussianDistribution();
            result.Mean = Mean;
            result.StandardDeviation = StandardDeviation;
            result.Variance = Variance;
            result.Precision = Precision;
            result.PrecisionMean = PrecisionMean;
            return result;
        }

        public static GaussianDistribution FromPrecisionMean(double precisionMean, double precision)
        {
            var gaussianDistribution = new GaussianDistribution();
            gaussianDistribution.Precision = precision;
            gaussianDistribution.PrecisionMean = precisionMean;
            gaussianDistribution.Variance = 1.0/precision;
            gaussianDistribution.StandardDeviation = Math.Sqrt(gaussianDistribution.Variance);
            gaussianDistribution.Mean = gaussianDistribution.PrecisionMean/gaussianDistribution.Precision;
            return gaussianDistribution;
        }

        // Although we could use equations from // For details, see http://www.tina-vision.net/tina-knoppix/tina-memo/2003-003.pdf
        // for multiplication, the precision mean ones are easier to write :)
        public static GaussianDistribution operator *(GaussianDistribution left, GaussianDistribution right)
        {
            return FromPrecisionMean(left.PrecisionMean + right.PrecisionMean, left.Precision + right.Precision);
        }

        /// Computes the absolute difference between two Gaussians
        public static double AbsoluteDifference(GaussianDistribution left, GaussianDistribution right)
        {
            return Math.Max(
                Math.Abs(left.PrecisionMean - right.PrecisionMean),
                Math.Sqrt(Math.Abs(left.Precision - right.Precision)));
        }

        /// Computes the absolute difference between two Gaussians
        public static double operator -(GaussianDistribution left, GaussianDistribution right)
        {
            return AbsoluteDifference(left, right);
        }

        public static double LogProductNormalization(GaussianDistribution left, GaussianDistribution right)
        {
            if ((left.Precision == 0) || (right.Precision == 0))
            {
                return 0;
            }

            double varianceSum = left.Variance + right.Variance;
            double meanDifference = left.Mean - right.Mean;

            double logSqrt2Pi = Math.Log(Math.Sqrt(2*Math.PI));
            return -logSqrt2Pi - (Math.Log(varianceSum)/2.0) - (Square(meanDifference)/(2.0*varianceSum));
        }


        public static GaussianDistribution operator /(GaussianDistribution numerator, GaussianDistribution denominator)
        {
            return FromPrecisionMean(numerator.PrecisionMean - denominator.PrecisionMean,
                                     numerator.Precision - denominator.Precision);
        }

        public static double LogRatioNormalization(GaussianDistribution numerator, GaussianDistribution denominator)
        {
            if ((numerator.Precision == 0) || (denominator.Precision == 0))
            {
                return 0;
            }

            double varianceDifference = denominator.Variance - numerator.Variance;
            double meanDifference = numerator.Mean - denominator.Mean;

            double logSqrt2Pi = Math.Log(Math.Sqrt(2*Math.PI));

            return Math.Log(denominator.Variance) + logSqrt2Pi - Math.Log(varianceDifference)/2.0 +
                   Square(meanDifference)/(2*varianceDifference);
        }

        private static double Square(double x)
        {
            return x*x;
        }

        public static double At(double x)
        {
            return At(x, 0, 1);
        }

        public static double At(double x, double mean, double standardDeviation)
        {
            // See http://mathworld.wolfram.com/NormalDistribution.html
            // 1 -(x-mean)^2 / (2*stdDev^2)
            // P(x) = ------------------- * e
            // stdDev * sqrt(2*pi)

            double multiplier = 1.0/(standardDeviation*Math.Sqrt(2*Math.PI));
            double expPart = Math.Exp((-1.0*Math.Pow(x - mean, 2.0))/(2*(standardDeviation*standardDeviation)));
            double result = multiplier*expPart;
            return result;
        }

        public static double CumulativeTo(double x, double mean, double standardDeviation)
        {
            double invsqrt2 = -0.707106781186547524400844362104;
            double result = ErrorFunctionCumulativeTo(invsqrt2*x);
            return 0.5*result;
        }

        public static double CumulativeTo(double x)
        {
            return CumulativeTo(x, 0, 1);
        }

        private static double ErrorFunctionCumulativeTo(double x)
        {
            // Derived from page 265 of Numerical Recipes 3rd Edition
            double z = Math.Abs(x);

            double t = 2.0/(2.0 + z);
            double ty = 4*t - 2;

            double[] coefficients = {
                                        -1.3026537197817094, 6.4196979235649026e-1,
                                        1.9476473204185836e-2, -9.561514786808631e-3, -9.46595344482036e-4,
                                        3.66839497852761e-4, 4.2523324806907e-5, -2.0278578112534e-5,
                                        -1.624290004647e-6, 1.303655835580e-6, 1.5626441722e-8, -8.5238095915e-8,
                                        6.529054439e-9, 5.059343495e-9, -9.91364156e-10, -2.27365122e-10,
                                        9.6467911e-11, 2.394038e-12, -6.886027e-12, 8.94487e-13, 3.13092e-13,
                                        -1.12708e-13, 3.81e-16, 7.106e-15, -1.523e-15, -9.4e-17, 1.21e-16, -2.8e-17
                                    };

            int ncof = coefficients.Length;
            double d = 0.0;
            double dd = 0.0;


            for (int j = ncof - 1; j > 0; j--)
            {
                double tmp = d;
                d = ty*d - dd + coefficients[j];
                dd = tmp;
            }

            double ans = t*Math.Exp(-z*z + 0.5*(coefficients[0] + ty*d) - dd);
            return x >= 0.0 ? ans : (2.0 - ans);
        }


        private static double InverseErrorFunctionCumulativeTo(double p)
        {
            // From page 265 of numerical recipes

            if (p >= 2.0)
            {
                return -100;
            }
            if (p <= 0.0)
            {
                return 100;
            }

            double pp = (p < 1.0) ? p : 2 - p;
            double t = Math.Sqrt(-2*Math.Log(pp/2.0)); // Initial guess
            double x = -0.70711*((2.30753 + t*0.27061)/(1.0 + t*(0.99229 + t*0.04481)) - t);

            for (int j = 0; j < 2; j++)
            {
                double err = ErrorFunctionCumulativeTo(x) - pp;
                x += err/(1.12837916709551257*Math.Exp(-(x*x)) - x*err); // Halley
            }

            return p < 1.0 ? x : -x;
        }

        public static double InverseCumulativeTo(double x, double mean, double standardDeviation)
        {
            // From numerical recipes, page 320
            return mean - Math.Sqrt(2)*standardDeviation*InverseErrorFunctionCumulativeTo(2*x);
        }

        public static double InverseCumulativeTo(double x)
        {
            return InverseCumulativeTo(x, 0, 1);
        }


        public override string ToString()
        {
            // Debug help
            return String.Format("μ={0:0.0000}, σ={1:0.0000}",
                                 Mean,
                                 StandardDeviation);
        }
    }
}
Something went wrong with that request. Please try again.