/// <include file='FactorDocs.xml' path='factor_docs/message_op_class[@name="BetaOp"]/message_doc[@name="MeanAverageLogarithm(Beta, Beta, Gamma, Beta)"]/*'/>
public static Beta MeanAverageLogarithm([Proper] Beta prob, Beta mean, [Proper] Gamma totalCount, Beta to_mean)
{
// Calculate gradient using method for DirichletOp
double ELogP, ELogOneMinusP;
prob.GetMeanLogs(out ELogP, out ELogOneMinusP);
Vector gradS = DirichletOp.CalculateGradientForMean(
Vector.FromArray(new double[] { mean.TrueCount, mean.FalseCount }),
totalCount,
Vector.FromArray(new double[] { ELogP, ELogOneMinusP }));
// Project onto a Beta distribution
Matrix A = new Matrix(2, 2);
double c = MMath.Trigamma(mean.TotalCount);
A[0, 0] = MMath.Trigamma(mean.TrueCount) - c;
A[1, 0] = A[0, 1] = -c;
A[1, 1] = MMath.Trigamma(mean.FalseCount) - c;
Vector theta = GammaFromShapeAndRateOp.twoByTwoInverse(A) * gradS;
Beta approximateFactor = new Beta(theta[0] + 1, theta[1] + 1);
if (damping == 0.0)
{
return(approximateFactor);
}
else
{
return((approximateFactor ^ (1 - damping)) * (to_mean ^ damping));
}
}
/// <include file='FactorDocs.xml' path='factor_docs/message_op_class[@name="LogOp_VMP"]/message_doc[@name="DAverageLogarithm(Gaussian, Gamma, Gamma)"]/*'/>
public static Gamma DAverageLogarithm([SkipIfUniform] Gaussian log, [SkipIfUniform] Gamma d, Gamma to_D)
{
if (log.IsPointMass)
{
return(DAverageLogarithm(log.Point, d, to_D));
}
Vector grad = Vector.Zero(2);
double meanLog = d.GetMeanLog();
double m, v;
log.GetMeanAndVariance(out m, out v);
grad[0] = -MMath.Tetragamma(d.Shape) / (2 * v) - MMath.Trigamma(d.Shape) / v * (meanLog - m);
grad[1] = (meanLog - m) / (v * d.Rate);
Gamma approximateFactor = GammaFromShapeAndRateOp.NonconjugateProjection(d, grad);
if (damping == 0.0)
{
return(approximateFactor);
}
else
{
return((approximateFactor ^ (1 - damping)) * (to_D ^ damping));
}
}