/// <summary> /// VMP message to 'probsTrue' /// </summary> /// <param name="sample">Incoming message from 'sample'. Must be a proper distribution. If any element is uniform, the result will be uniform.</param> /// <returns>The outgoing VMP message to the 'probsTrue' argument</returns> /// <remarks><para> /// The outgoing message is the exponential of the average log-factor value, where the average is over all arguments except 'probsTrue'. /// The formula is <c>exp(sum_(sample) p(sample) log(factor(sample,probsTrue)))</c>. /// </para></remarks> /// <exception cref="ImproperMessageException"><paramref name="sample"/> is not a proper distribution</exception> public static SparseBetaList ProbsTrueAverageLogarithm([Proper] SparseBernoulliListBase sample) { // E[x*log(p) + (1-x)*log(1-p)] = E[x]*log(p) + (1-E[x])*log(1-p) SparseVector ex = sample.GetProbTrueVector(); return(new SparseBetaList((SparseVector)(ex + 1), (SparseVector)(2 - ex))); }
/// <summary> /// The maximum 'difference' between this instance and that instance. /// This returns the maximum absolute difference between the Log-odds of any element /// </summary> /// <param name="that">The other distribution</param> /// <returns>The resulting maximum difference</returns> /// <remarks><c>a.MaxDiff(b) == b.MaxDiff(a)</c></remarks> public double MaxDiff(object that) { SparseBernoulliListBase sbl = (SparseBernoulliListBase)that; // differences in log odds SparseVector absDiff = SparseVector.Zero(LogOddsVector.Count); absDiff.SetToFunction(LogOddsVector, sbl.LogOddsVector, (a, b) => MMath.AbsDiff(a, b)); return(absDiff.Max()); }
//-- VMP ------------------------------------------------------------------------------------------- /// <summary> /// Evidence message for VMP /// </summary> /// <param name="sample">Incoming message from 'sample'. Must be a proper distribution. If any element is uniform, the result will be uniform.</param> /// <param name="probsTrue">Incoming message from 'probsTrue'. Must be a proper distribution. If any element is uniform, the result will be uniform.</param> /// <param name="MeanLog">Buffer 'MeanLog'.</param> /// <param name="MeanLogOneMinus">Buffer 'MeanLogOneMinus'.</param> /// <returns>Average of the factor's log-value across the given argument distributions</returns> /// <remarks><para> /// The formula for the result is <c>sum_(sample,probsTrue) p(sample,probsTrue) log(factor(sample,probsTrue))</c>. /// Adding up these values across all factors and variables gives the log-evidence estimate for VMP. /// </para></remarks> /// <exception cref="ImproperMessageException"><paramref name="sample"/> is not a proper distribution</exception> /// <exception cref="ImproperMessageException"><paramref name="probsTrue"/> is not a proper distribution</exception> public static double AverageLogFactor([Proper] SparseBernoulliListBase sample, [Proper] SparseBetaList probsTrue, SparseVector MeanLog, SparseVector MeanLogOneMinus) { //var MeanLogOneMinus = probsTrue.GetMeanLogOneMinus(); var p = sample.GetProbTrueVector(); var res = p * MeanLog + (1 - p) * MeanLogOneMinus; return(res.Sum()); }
/// <summary> /// Evidence message for VMP /// </summary> /// <param name="sample">Incoming message from 'sample'. Must be a proper distribution. If any element is uniform, the result will be uniform.</param> /// <param name="probsTrue">Constant value for 'probsTrue'.</param> /// <param name="MeanLog">Buffer 'MeanLog'.</param> /// <param name="MeanLogOneMinus">Buffer 'MeanLogOneMinus'.</param> /// <returns>Average of the factor's log-value across the given argument distributions</returns> /// <remarks><para> /// The formula for the result is <c>sum_(sample) p(sample) log(factor(sample,probsTrue))</c>. /// Adding up these values across all factors and variables gives the log-evidence estimate for VMP. /// </para></remarks> /// <exception cref="ImproperMessageException"><paramref name="sample"/> is not a proper distribution</exception> public static double AverageLogFactor([Proper] SparseBernoulliListBase sample, SparseVector probsTrue, SparseVector MeanLog, SparseVector MeanLogOneMinus) { return(AverageLogFactor(sample, SparseBetaList.PointMass(probsTrue), MeanLog, MeanLogOneMinus)); }