/// <summary>
/// Gets a vector of log P(false) values.
/// </summary>
/// <returns></returns>
public SparseVector GetLogProbFalseVector()
{
SparseVector sv = SparseVector.Zero(Count);
sv.SetToFunction(LogOddsVector, x => MMath.LogisticLn(-x));
return(sv);
}
/// <summary>Evidence message for VMP.</summary>
/// <param name="sample">Constant value for <c>sample</c>.</param>
/// <param name="logOdds">Constant value for <c>logOdds</c>.</param>
/// <returns>Average of the factor's log-value across the given argument distributions.</returns>
/// <remarks>
/// <para>The formula for the result is <c>log(factor(sample,logOdds))</c>. Adding up these values across all factors and variables gives the log-evidence estimate for VMP.</para>
/// </remarks>
public static double AverageLogFactor(bool sample, double logOdds)
{
if (sample)
{
return(MMath.LogisticLn(logOdds));
}
else
{
return(MMath.LogisticLn(-logOdds));
}
}
/// <summary>
/// Evidence message for VMP.
/// </summary>
/// <param name="sample">Incoming message from sample</param>
/// <param name="logOdds">Incoming message from logOdds</param>
/// <returns><c>sum_x marginal(x)*log(factor(x))</c></returns>
/// <remarks><para>
/// The formula for the result is <c>int log(f(x)) q(x) dx</c>
/// where <c>x = (sample,logOdds)</c>.
/// </para></remarks>
public static double AverageLogFactor(Bernoulli sample, [Proper, SkipIfUniform] Gaussian logOdds)
{
if (logOdds.IsUniform())
{
return(0.0);
}
double m, v;
logOdds.GetMeanAndVariance(out m, out v);
double t = Math.Sqrt(m * m + v);
double s = 2 * sample.GetProbTrue() - 1; // probTrue - probFalse
return(MMath.LogisticLn(t) + (s * m - t) / 2);
}
/// <summary>
/// Evidence message for VMP.
/// </summary>
/// <param name="sample">Fixed value for sample</param>
/// <param name="logOdds">Incoming message from logOdds</param>
/// <returns><c>sum_x marginal(x)*log(factor(x))</c></returns>
/// <remarks><para>
/// The formula for the result is <c>int log(f(x)) q(x) dx</c>
/// where <c>x = (sample,logOdds)</c>.
/// </para></remarks>
public static double AverageLogFactor(bool sample, [Proper, SkipIfUniform] Gaussian logOdds)
{
if (logOdds.IsUniform())
{
return(0.0);
}
double m, v;
logOdds.GetMeanAndVariance(out m, out v);
double t = Math.Sqrt(m * m + v);
double s = sample ? 1 : -1;
return(MMath.LogisticLn(t) + (s * m - t) / 2);
}
/// <summary>
/// Gets the log probability of the given value under this distribution
/// </summary>
/// <param name="value"></param>
/// <returns></returns>
public double GetLogProb(IList <int> value)
{
SparseVector tv = SparseVector.Zero(LogOddsVector.Count);
tv.SetToFunction(LogOddsVector, x => MMath.LogisticLn(-x));
double logprob = tv.Sum();
foreach (int index in value)
{
double logOdds = LogOddsVector[index];
logprob -= MMath.LogisticLn(-logOdds);
logprob += MMath.LogisticLn(logOdds);
}
return(logprob);
}
/// <summary>
/// VMP message to 'logistic'
/// </summary>
/// <param name="x">Incoming message from 'x'. Must be a proper distribution. If uniform, the result will be uniform.</param>
/// <returns>The outgoing VMP message to the 'logistic' argument</returns>
/// <remarks><para>
/// The outgoing message is a distribution matching the moments of 'logistic' as the random arguments are varied.
/// The formula is <c>proj[sum_(x) p(x) factor(logistic,x)]</c>.
/// </para></remarks>
/// <exception cref="ImproperMessageException"><paramref name="x"/> is not a proper distribution</exception>
public static Beta LogisticAverageLogarithm([Proper] Gaussian x)
{
double m, v;
x.GetMeanAndVariance(out m, out v);
// for consistency with XAverageLogarithm
double t = Math.Sqrt(m * m + v);
double s = -1;
double eLogOneMinusP = MMath.LogisticLn(t) + (s * m - t) / 2;
// E[log (1-sigma(x))] = E[log sigma(-x)] = -E[log(1+exp(x))]
// E[log sigma(x)] = -E[log(1+exp(-x))] = -E[log(1+exp(x))-x] = -E[log(1+exp(x))] + E[x]
double eLogP = eLogOneMinusP + m;
return(Beta.FromMeanLogs(eLogP, eLogOneMinusP));
}
/// <summary>
/// Evidence message for VMP.
/// </summary>
/// <param name="sample">Incoming message from sample</param>
/// <param name="logOdds">Incoming message from logOdds</param>
/// <returns><c>sum_x marginal(x)*log(factor(x))</c></returns>
/// <remarks><para>
/// The formula for the result is <c>int log(f(x)) q(x) dx</c>
/// where <c>x = (sample,logOdds)</c>.
/// </para></remarks>
public static double AverageLogFactor(Beta logistic, [Proper, SkipIfUniform] Gaussian x, Beta to_logistic)
{
double a = logistic.TrueCount;
double b = logistic.FalseCount;
double scale = a + b - 2;
double shift = -(b - 1);
// sigma(x) >= sigma(t) exp((x-t)/2 - a/2*(x^2 - t^2))
double m, v;
x.GetMeanAndVariance(out m, out v);
double t = Math.Sqrt(m * m + v);
double lambda = (t == 0) ? 0.25 : Math.Tanh(t / 2) / (2 * t);
double boundOnLogSigma = MMath.LogisticLn(t) + (m - t) / 2.0 - .5 * lambda * (m * m + v - t * t);
return(scale * boundOnLogSigma + shift * m - logistic.GetLogNormalizer() - to_logistic.GetAverageLog(logistic));
}
public double GetLogProb(int value)
{
if (value < 0 || value > TrialCount)
{
return(double.NegativeInfinity);
}
if (IsPointMass)
{
return((value == Point) ? 1.0 : 0.0);
}
return(MMath.GammaLn(TrialCount + 1) - A * MMath.GammaLn(value + 1) - B * MMath.GammaLn(TrialCount - value + 1) + value * LogOdds + TrialCount * MMath.LogisticLn(-LogOdds));
}
/// <summary>
/// Gets the log probability of the binary variable being true
/// </summary>
/// <returns>log(p(x=true))</returns>
public double GetLogProbTrue()
{
return(MMath.LogisticLn(LogOdds));
}
/// <summary>
/// Computes the log-probability that A==B where p(A)=Logistic(x), p(B)=Logistic(y).
/// </summary>
/// <param name="x">The log-odds of variable A, which can be any real number from -Inf to Inf.</param>
/// <param name="y">The log-odds of variable B, which can be any real number from -Inf to Inf.</param>
/// <returns>The log-probability that A==B.</returns>
/// <remarks>
/// The result is Math.Log(0.5) if x=0 or y=0.
/// </remarks>
public static double LogProbEqual(double x, double y)
{
return(MMath.LogisticLn(LogitProbEqual(x, y)));
}
/// <summary>
/// Evaluates the logarithm of the density function
/// </summary>
/// <param name="x">true or false</param>
/// <returns>Log of the probability density for the given event</returns>
public double GetLogProb(bool x)
{
return(MMath.LogisticLn(x ? LogOdds : -LogOdds));
}
/// <summary>
/// Gets the log probability of the binary variable being false
/// </summary>
/// <returns>log(p(x=false))</returns>
public double GetLogProbFalse()
{
return(MMath.LogisticLn(-LogOdds));
}
/// <summary>Evidence message for EP.</summary>
/// <param name="sample">Constant value for <c>sample</c>.</param>
/// <param name="logOdds">Constant value for <c>logOdds</c>.</param>
/// <returns>Logarithm of the factor's average value across the given argument distributions.</returns>
/// <remarks>
/// <para>The formula for the result is <c>log(factor(sample,logOdds))</c>.</para>
/// </remarks>
public static double LogAverageFactor(bool sample, double logOdds)
{
return(MMath.LogisticLn(sample ? logOdds : -logOdds));
}
