API
Models
BayesianQuadrature.BayesModel
— TypeBayesModel(prior, logintegrand) <: AbstractBQModel
Model inheriting from AbstractMCMC.AbstractModel
. prior
should be a multivariate distribution from Distributions.jl
at the moment prior
has to be a MvNormal
but this will improved in a later version logintegrand
should be the log of the function to integrate.
Samplers
BayesianQuadrature.PriorSampling
— TypePriorSampling()
Sampler which will use the prior distribution from the given model to provide samples.
Bayesian Quadratures
BayesianQuadrature.BayesQuad
— TypeBayesQuad(k::Kernel; l=1.0, σ::Real=1.0)
Tool for estimating the bayesian quadrature for a given BayesQuadModel
and a Sampler
.
BayesQuad
estimate the probability distribution p(I)
of I = ∫ f(x) p(x) dx
Argument
k::Kernel
: kernel fromKernelFunctions.jl
, the kernel cannot be composite
Keywords argument
l
: lengthscale of the kernel. It can be:Real
: isotropic kernelAbstractVector
: ARD kernel (one lengthscale per dimension)LowerTriangular
: Linear transformation of the inputs
σ
: variance of the kernel (k(x,x') -> σ * k(x,x')
)
Note: If k
already has a variance and/or a transformation, these will be automatically extracted and replace the given keyword arguments