low-rank basis matrix, (outputdim*numberOfPoints)x(rank)
scaled basis matrix, scaled with standard deviations ("W" [PPCA lit.] or "Q" [GP lit.])
Get the low-rank expansion coefficients of a sample, respects model noise
Get the probability distribution of the coefficients in the latent space of the model (multivariate standard normal).
Get the probability distribution of the coefficients in the latent space of the model (multivariate standard normal).
MultivariateNormalDistribution(0,I).
covariance of this Gaussian Process: underlying GP + local uncorrelated noise
domain where this Gaussian Process is defined (same as underling process)
underlying discrete low-rank Gaussian Process, subspace model
Draw an instance from the model, instance of underlying DiscreteLowRankGaussianProcess
Get instance at a specific point
interpolate this process with nearest neighbours
Returns the log of the probability density of the instance
Returns the log of the probability density of the instance
If you are interested in ordinal comparisons of PDFs, use this as it is numerically more stable, takes spherical noise term into account
Calculate the marginal distribution on a set of points
inverted intermediate matrix M
mean field
mean vector of underlying DLRGP model
Noise distribution, expands the low rank subspace distribution to whole domain/space
isotropic noise variance, "subspace + N(0, noiseVariance)"
output dimension of GP, same as underlying
Returns the probability density of the given instance, takes spherical noise term into account
Calculate the posterior model given point observations with an individual uncertainty for each
Calculate the posterior model given point observations with an individual uncertainty for each
list of point observations (PointId, Value, Uncertainty), uncertainty is *additional* (independent) to model noise
Calculate the posterior model given point observations with a common independent isotropic Gaussian noise
Calculate the posterior model given point observations with a common independent isotropic Gaussian noise
observation noise of sample, *additional* (independent) to model noise
Project the sample into this model, best reconstruction
See DiscreteLowRankGaussianProcess.rank
Draw a sample from the model, includes noise(!) (use underlying gpModel without noise)
standard deviation along each basis direction
this process as a DiscreteGaussianProcess, full rank
truncate the underlying low rank GP model (drops components, does not change noise)
variance along each basis direction
(Since version ) see corresponding Javadoc for more information.
Models of the type DiscreteLowRankGaussianProcess + noiseVariance * delta - extension of DLRGP to whole space (pancake) This is a PPCA model but the basis matrix does not need to be orthogonal
dimensionality of domain
value type of model
underlying discrete low-rank Gaussian Process, subspace model
isotropic noise variance, "subspace + N(0, noiseVariance)"