public static double mlfHelperNode(PositiveDefiniteMatrix prec, PositiveDefiniteMatrix[] gradK, Vector gradientVector, Vector[] functions_SampleMean,
PositiveDefiniteMatrix[] functions_SampleVariance,
DistributionStructArray <Gamma, double> nodeSignalPrecisions)
{
var logDetPrec = VectorGaussianScaledPrecisionOp.PrecisionMeanLogDet(prec);
double res = 0;
if (gradientVector != null)
{
for (int i = 0; i < gradientVector.Count; i++)
{
gradientVector[i] = 0;
}
}
for (int q = 0; q < functions_SampleMean.Length; q++)
{
res += VectorGaussianScaledPrecisionOp.AverageLogFactor(functions_SampleMean[q], nodeSignalPrecisions[q],
VectorGaussianScaledPrecisionOp.ESamplePrecisionSample(functions_SampleMean[q], functions_SampleVariance[q], prec),
logDetPrec);
if (gradientVector != null)
{
for (int i = 0; i < gradientVector.Count; i++)
{
gradientVector[i] += VectorGaussianScaledPrecisionOp.GradAverageLogFactor(
functions_SampleMean[q],
functions_SampleVariance[q],
nodeSignalPrecisions[q],
gradK[i],
prec);
}
}
}
return(res);
}
public static double AverageLogFactor(
Vector SampleMean,
double scaling,
GPBuffer Buffer)
{
return(VectorGaussianScaledPrecisionOp.AverageLogFactor(SampleMean, scaling, Buffer.ESamplePrecisionSample, Buffer.PrecisionMeanLogDet));
}
public static double mlfHelperWeight(PositiveDefiniteMatrix prec, PositiveDefiniteMatrix[] gradK, Vector gradientVector, Vector[][] functions_SampleMean,
PositiveDefiniteMatrix[][] functions_SampleVariance)
{
double res = 0;
var logDetPrec = VectorGaussianScaledPrecisionOp.PrecisionMeanLogDet(prec);
if (gradientVector != null)
{
for (int i = 0; i < gradientVector.Count; i++)
{
gradientVector[i] = 0;
}
}
int D = functions_SampleMean[0].Length;
var resArray = new double[D];
Vector[] gradVArray = gradientVector != null?System.Linq.Enumerable.Range(0, D).Select(_ => DenseVector.Zero(gradientVector.Count)).ToArray() : null;
for (int q = 0; q < functions_SampleMean.Length; q++)
{ //for (int d = 0; d < functions_SampleMean[0].Length; d++)
Parallel.For(0, D, d =>
{
resArray[d] = VectorGaussianScaledPrecisionOp.AverageLogFactor(functions_SampleMean[q][d], 1,
VectorGaussianScaledPrecisionOp.ESamplePrecisionSample(functions_SampleMean[q][d], functions_SampleVariance[q][d], prec),
logDetPrec);
if (gradientVector != null)
{
for (int i = 0; i < gradientVector.Count; i++)
{
gradVArray[d][i] = VectorGaussianScaledPrecisionOp.GradAverageLogFactor(
functions_SampleMean[q][d],
functions_SampleVariance[q][d],
Gamma.PointMass(1),
gradK[i],
prec);
}
}
});
}
res = resArray.Sum();
if (gradientVector != null)
{
gradientVector.SetTo(gradVArray.Aggregate((p, q) => p + q));
}
return(res);
}
/// <summary>
/// Uses the KernelOptimiser class to optimise the hypers given the current variational posterior
/// on the function values (which has mean SampleMean and covariance SampleCovariance)
/// </summary>
public static GPBuffer BufferHelper(int[] hypersToOptimise, GPBuffer Buffer, Vector[] x, Vector SampleMean, PositiveDefiniteMatrix SampleVariance, Gamma scaling)
{
if (SampleMean.All(o => o == 0.0))
{
Buffer.Precision = Utils.GramMatrix(Buffer.kernel, x).Inverse();
}
else
{
//Console.WriteLine(Utils.KernelToArray(Buffer.kernel).Select(o => o.ToString()).Aggregate((p, q) => p + " " + q));
var helper = new KernelOptimiser(settings);
helper.kernel = Buffer.kernel;
helper.xData = x;
helper.hypersToOptimise = hypersToOptimise;
helper.Optimise((prec, gradK, gradientVector) =>
helperFunction(prec, gradK, gradientVector, scaling, SampleMean,
SampleVariance), ref Buffer.Precision);
Buffer.ESamplePrecisionSample = VectorGaussianScaledPrecisionOp.ESamplePrecisionSample(SampleMean, SampleVariance, Buffer.Precision);
Buffer.PrecisionMeanLogDet = VectorGaussianScaledPrecisionOp.PrecisionMeanLogDet(Buffer.Precision);
//Console.WriteLine(Utils.KernelToArray(Buffer.kernel).Select(o => o.ToString()).Aggregate((p, q) => p + " " + q));
rememberKernel = Buffer.kernel;
}
return(Buffer);
}
/// <summary>
/// Calculates the variational lower bound and its derivatives wrt to the hypers. gradK is the gradient of the kernel matrix
/// wrt to the hypers.
/// </summary>
public static double helperFunction(PositiveDefiniteMatrix prec, PositiveDefiniteMatrix[] gradK, Vector gradientVector, Gamma scaling,
Vector functions_SampleMean,
PositiveDefiniteMatrix functions_SampleVariance)
{
double res = 0;
var logDetPrec = VectorGaussianScaledPrecisionOp.PrecisionMeanLogDet(prec);
res = VectorGaussianScaledPrecisionOp.AverageLogFactor(functions_SampleMean, scaling,
VectorGaussianScaledPrecisionOp.ESamplePrecisionSample(functions_SampleMean, functions_SampleVariance, prec),
logDetPrec);
if (gradientVector != null)
{
for (int i = 0; i < gradientVector.Count; i++)
{
gradientVector[i] = VectorGaussianScaledPrecisionOp.GradAverageLogFactor(
functions_SampleMean,
functions_SampleVariance,
Gamma.PointMass(1),
gradK[i],
prec);
}
}
return(res);
}
