/// <include file='FactorDocs.xml' path='factor_docs/message_op_class[@name="SumVectorGaussianOp"]/message_doc[@name="ArrayAverageConditional{TVectorGaussianList}(VectorGaussian, VectorGaussian, IList{VectorGaussian}, TVectorGaussianList)"]/*'/>
public static TVectorGaussianList ArrayAverageConditional <TVectorGaussianList>(
[SkipIfUniform] VectorGaussian sum,
[Fresh] VectorGaussian to_sum,
IList <VectorGaussian> array,
TVectorGaussianList result)
where TVectorGaussianList : IList <VectorGaussian>, SettableToUniform
{
// Check inputs for consistency
int dimension = CheckArgumentConsistency(sum, to_sum, array, result);
if (array.Count == 0)
{
return(result);
}
// It is tempting to put SkipIfAllUniform on array but this isn't correct if the array has one element
if (array.Count == 1)
{
result[0].SetTo(sum);
return(result);
}
if (!sum.IsProper())
{
foreach (VectorGaussian element in result)
{
element.SetTo(sum);
}
return(result);
}
var elementMean = Vector.Zero(dimension);
var elementVariance = PositiveDefiniteMatrix.Identity(dimension);
// Check if an element of the array is uniform
int indexOfUniform = -1;
for (int i = 0; i < array.Count; i++)
{
array[i].GetMeanAndVariance(elementMean, elementVariance);
// Instead of testing IsUniform, we need to test the more strict requirement that all diagonal
// elements of variance are infinite due to the way we are doing the computations
if (IsUniform(elementVariance))
{
if (indexOfUniform >= 0)
{
// More than one element of array is uniform
result.SetToUniform();
return(result);
}
indexOfUniform = i;
}
}
Vector sumMean = Vector.Zero(dimension);
PositiveDefiniteMatrix sumVariance = PositiveDefiniteMatrix.Identity(dimension);
sum.GetMeanAndVariance(sumMean, sumVariance);
Vector totalMean = Vector.Zero(sum.Dimension);
PositiveDefiniteMatrix totalVariance = PositiveDefiniteMatrix.IdentityScaledBy(sum.Dimension, 0.0);
if (indexOfUniform >= 0)
{
// Exactly one element of array is uniform
for (int i = 0; i < array.Count; i++)
{
if (i == indexOfUniform)
{
continue;
}
array[i].GetMeanAndVariance(elementMean, elementVariance);
totalMean.SetToSum(totalMean, elementMean);
totalVariance.SetToSum(totalVariance, elementVariance);
result[i].SetToUniform();
}
// totalMean = sum_{i except indexOfUniform} array[i].GetMean()
// totalVariance = sum_{i except indexOfUniform} array[i].GetVariance()
totalMean.SetToDifference(sumMean, totalMean);
totalVariance.SetToSum(sumVariance, totalVariance);
result[indexOfUniform].SetMeanAndVariance(totalMean, totalVariance);
return(result);
}
// At this point, the array has no uniform elements
// Get the mean and variance of sum of all Gaussians
to_sum.GetMeanAndVariance(totalMean, totalVariance);
// Subtract it off from the mean and variance of the incoming Gaussian from Sum
totalMean.SetToDifference(sumMean, totalMean);
totalVariance.SetToSum(totalVariance, sumVariance);
for (int i = 0; i < array.Count; i++)
{
array[i].GetMeanAndVariance(elementMean, elementVariance);
elementMean.SetToSum(elementMean, totalMean);
elementVariance.SetToDifference(totalVariance, elementVariance);
result[i].SetMeanAndVariance(elementMean, elementVariance);
}
return(result);
}
